multiplayer roguelikes, surreal time, etc.
crichmon
I've been searching for anything on multiplayer roguelikes. Unfortunately,
a lot of the links aren't valid anymore. I was able to download some of the
source for Interhack... can anyone help on sources and webpages?
What do you think of a multi-player roguelike?
What are some of the design issues?
What is "surreal time"? The documentation that I found had all text removed
that refered to the topic.
crichmon
Brent Ross
crichmon <crich...@hotmail.com> wrote:
// What do you think of a multi-player roguelike?
Better done by other genres of game.
// What are some of the design issues?
You need to move towards a real time system, since if you don't
"time out" a player then they might end up holding up everyone else.
On the other hand, the complexity of roguelikes often makes it better
if the player gets to think about their moves a bit (or at least the
opportunty to comfortably navigate menus and targeting systems), thus the
time out has to be reasonably large or the game made relatively simple.
Walking down long corridors with a timeout of a second can be painful...
as can melee (learnt this in Mangband... also learnt that it was easiest
just to stand in the middle of a room of breeders and let the automatic
strikebacks do the killing, and wait to gain levels... walking down
another long corridor to get to another room wasn't a fun alternative to
letting the character farm his way up). In short, most roguelikes are
designed to take advantage of the fact that the player can play very fast,
and then suddenly very slow (taking 5 minutes when you're a turn a way
from death)... so if you're going to break the turn system you might as
well start thinking from scratch (certainly, you'll want a much simpler
than average RL game with a simpler interface). That's why I think of
multiplayer as best for other genres.
// What is "surreal time"? The documentation that I found had all text removed
// that refered to the topic.
Sureal time, as I remember, was a system where the game worked as a single
player game until two PCs got within each others "sphere of influence",
at which point the game would move into a real time mode. It was an
attempt at a compromise. The name "sureal time" is very apt... the
PCs have a bizarre relative effect that happens based on the speed of
their players (two characters who meet again in the future might have
radically different amounts of time pass... something which is related
to the players' relative speed of play).
Brent Ross
Ray Dillinger
> Hi,
>
> I've been searching for anything on multiplayer roguelikes. Unfortunately,
> a lot of the links aren't valid anymore. I was able to download some of the
> source for Interhack... can anyone help on sources and webpages?
>
> What do you think of a multi-player roguelike?
It is interesting, but not easy to do well. There are very tough
design decisions you will have to make.
> What are some of the design issues?
There is really only one. The design issue is time.
Roguelikes are turn-based. That means that if you
feel like you're getting stupid in the middle of a
pitched battle, you can get up and go mix yourself
a drink and calm your nerves and have dinner with
your family while you ponder what you ought to do
next, and when you get back, you still have that
moment of time in which you can do whatever you
decided.
Now, imagine that you have a multi-player roguelike
and one character stops the passage of time because
his player wants to think about the next move, go
to the bathroom, or not be late for a date with his
girlfriend. What happens to all the other players?
Are they sitting there, tearing their hair out, while
their characters are stuck between ticks of the
clock? Not fun.
So, that's what happens when dungeon time proceeds at
the rate of the slowest player. But time proceeding
at the rate of the fastest player isn't much better;
in that case players miss actions because they didn't
hit the buttons as fast as machine-gun mike, and the
monsters all move at mike's speed, and if anybody
pauses for breath, he gets creamed.
Somewhere in the middle, to make a multiplayer roguelike,
you have to strike a compromise which most of your
players will bitch and moan about.
Your choices are:
1) absolute-time timeouts. This is "if I don't get
your move in N milliseconds, your character
doesn't act."
2) relative-time timeouts. This is "if I don't get
your move within N milliseconds after getting the
first/second/third move taken this round, your
character doesn't act. A relative-timeout to the
third action means that if less than three people
send in an action, time is stopped until the third
person sends one in - and at that point there are N
milliseconds for everybody else's action to come in.
3) Timeouts with a halt action. A player who needs
to browse menus, etc, may send a "Halt," which
stops time dead, including for the monsters and
other players, to contemplate their action and then
time starts again when the action arrives. But if
a player doesn't send "halt" they are subject to
absolute or relative timeouts, as above.
4) Timeouts with a limited halt action. As above, but
time restarts N milliseconds later regardless.
5) Surreal time. Time need not always pass for all
characters at the same rate.
> What is "surreal time"? The documentation that I found had all text removed
> that refered to the topic.
Surreal time allowed players to separate their characters
in the dungeon and then proceed at independent rates of
dungeon time. But when they ran into each other again,
turn-based with timeouts was the rule again. This meant
that your character could get separated from the party
for a few minutes of their time and you could run him
through several hours of his time.
Bear
Glen Wheeler
"Brent Ross" <bwr...@csclub.uwaterloo.ca> wrote in message
news:cksgto$jba$1...@rumours.uwaterloo.ca...
This is *very* interesting; although could be quite unnerving. Does
anybody know if MAngband plans on including this?
--
Glen
L:Pyt E+++ T-- R+ P+++ D+ G+ F:*band !RL RLA-
W:AF Q+++ AI++ GFX++ SFX-- RN++++ PO--- !Hp Re-- S+
Brent Ross
Ray Dillinger <be...@sonic.net> wrote:
// at the rate of the fastest player isn't much better;
// in that case players miss actions because they didn't
// hit the buttons as fast as machine-gun mike, and the
// monsters all move at mike's speed, and if anybody
// pauses for breath, he gets creamed.
There is a sixth option that can lessen machine-gun Mike's advantage...
namely, allow N actions per time out. With this, MGM has his advantage
capped, while not feeling constantly hindered by long timeouts for every
action... and slow players can relax a bit knowing that the game isn't
going to race totally out of control on them. It's a bit hard to tune,
and really is a compromise for both sides (slow players need to rush a
bit, fast players experience some jerkiness)... but it is an option.
Brent Ross
Ray Dillinger
I wonder if there's some version of "surreal time" that can be
applied to closely interacting characters usefully. Really,
the thing is about how fast the monsters move. If MGM over
there is using up ten game minutes per wall minute, and you're
in the other corner of the room using up two game minutes per
wall minute, how fast are the monsters in the room supposed
to act? Maybe the ones closer to MGM should move at closer
to his speed; Maybe the ones that are attacking/defending from
MGM should move closer to his speed.
You could do something where actions missed due to timeouts
go into a sort of "reserve" that the party shares, and then
if MGM gets two actions or more during a single timeout period
the "extra" actions come out of the reserve. That way you
wouldn't have the monsters driven way faster because one
character is pressing keys fast. The monsters speeds would
sort of "average out" to the rate at which the entire party
took turns. But this business of transferring turns from
one player to another is fraught with opportunities for
heinous abuse.
Bear
crichmon
> crichmon wrote:
>
>> I've been searching for anything on multiplayer
>> roguelikes. Unfortunately, a lot of the links
>> aren't valid anymore. I was able to download some
>> of the source for Interhack... can anyone help on
>> sources and webpages?
>>
>> What do you think of a multi-player roguelike?
>
> It is interesting, but not easy to do well. There
> are very tough design decisions you will have to
> make.
>
>> What are some of the design issues?
>
> There is really only one. The design issue is time.
> Roguelikes are turn-based.
> that's what happens when dungeon time proceeds at
> the rate of the slowest player. But time proceeding
> at the rate of the fastest player isn't much better;
> Somewhere in the middle, to make a multiplayer
> roguelike, you have to strike a compromise which most
> of your players will bitch and moan about.
Okay, so a multi-player roguelike requires the roguelike to handle time in
more of a real-time fashion.
This makes me ask a related tangent: okay, what do you think about a
single-player roguelike that uses a time-system other then turn-based (e.g.,
real-time)?
I see. When I had thought about a multi-player roguelike in the past, I was
considering a system similar to #1 or #2.
I thought perhaps one possible project path would be to make a single-player
somewhat real-time roguelike, and then after that modify it to make it
multi-player. What do you think of that idea? Elements could be added to
the game to allow the player an amount of time to deal actions that are more
complex and less immediate-action, such as a party of adventurers securing a
dungeon camp.
crichmon
Brent Ross
// Brent Ross wrote:
// > There is a sixth option that can lessen machine-gun Mike's advantage...
// > namely, allow N actions per time out. With this, MGM has his advantage
// > capped, while not feeling constantly hindered by long timeouts for every
// > action... and slow players can relax a bit knowing that the game isn't
// > going to race totally out of control on them. It's a bit hard to tune,
// > and really is a compromise for both sides (slow players need to rush a
// > bit, fast players experience some jerkiness)... but it is an option.
// >
//
// I wonder if there's some version of "surreal time" that can be
// applied to closely interacting characters usefully. Really,
// the thing is about how fast the monsters move. If MGM over
// there is using up ten game minutes per wall minute, and you're
// in the other corner of the room using up two game minutes per
// wall minute, how fast are the monsters in the room supposed
// to act? Maybe the ones closer to MGM should move at closer
// to his speed; Maybe the ones that are attacking/defending from
// MGM should move closer to his speed.
I think the problem with that is that it's probably too surreal. Also,
how do you tell which monsters are in who's effect radius. It could
easily be that most of the time the monsters and players have such huge
overlap that it never can be easily split. For example, consider spells
that attack all monsters in LoS (eg Dispel Evil)... if MGM is playing
a character with one, the monsters in the corner pounding slow-poke Sue
in melee should get to react to him (thus they get sucked into his time
frame and SPS probably suffers), or else the players can abuse surreal
time to slow monsters.
Brent Ross
Lauri Vallo
>I thought perhaps one possible project path would be to make a single-player
>somewhat real-time roguelike, and then after that modify it to make it
>multi-player. What do you think of that idea? Elements could be added to
>the game to allow the player an amount of time to deal actions that are more
>complex and less immediate-action, such as a party of adventurers securing a
>dungeon camp.
This makes me think that the actions of your foes should change to be
reactions, so that you will never be attacked by hostile things if you
stay still.
crichmon
So while the player is acting, real-time takes place, and while the player
is not acting, or hasn't chosen an action, then the game time would have
stopped? Neat. I'm picturing some cool combat freeze-frames. That is
definitely one option.
However, I'm not sure if that (perhaps 'auto-pause'?) isn't as far away of a
step from the RL-mold that I'm interested in. My one thought is that if a
single-player rl _is_ real-time, then the effort to make it multi-player
wouldn't have the extra design issue.
crichmon
Gerry Quinn
crich...@hotmail.com says...
> This makes me ask a related tangent: okay, what do you think about a
> single-player roguelike that uses a time-system other then turn-based (e.g.,
> real-time)?
It's called Diablo and is quite popular...
- Gerry Quinn
Lorenzo Gatti
[...]
> Okay, so a multi-player roguelike requires the roguelike to handle time in
> more of a real-time fashion.
>
> This makes me ask a related tangent: okay, what do you think about a
> single-player roguelike that uses a time-system other then turn-based (e.g.,
> real-time)?
>
3059 is realtime; people move around and time passes while the game
waits for user input, with a fine granularity (more or less 1 second)
that makes timing matter: while you decide to talk with a character he
could have moved away.
I set aside 3059 because I want time to think, especially during my
first experience with a new game when I am necessarily slow because I
have to learn everything.
(3059 is an otherwise refined, original and interesting game; I just
dislike the choice of real time interaction.)
A multiplayer 3059-like game would be even worse: players would try to
be faster than other players to get an advantage, not only acceptably
fast for what they are doing.
With short intervals for actions, shooting and dodging "twitch" action
is only a few unwise rules away.
Lorenzo Gatti
ga...@dsdata.it
Pfhoenix
Whenever the player performs an action that takes time (to include using a
computer, i.e. entering commands and such, if there's another entity on the
same computer system/network, such as a security AI or anti-intrusion/tracer
module), the following happens :
* all loaded maps are checked to see if any entity on them is flagged as
Important (players are flagged as Important, as will be, later on down the
line, special or story important NPCs)
* all entities on the check-passed maps are given the chance to decide what
they want to do within the given time frame (however long it takes the
player to perform his/her action)
* a list is made of all entities that have a pending action, ordered by how
long it takes them to perform their action
* sequentially, each entity performs their action. As an entity acts, it
gets the chance to decide if it wants to act again, and if it decides to act
again (given that it has enough time left to act in), it gets re-sorted into
the list
* once all entities have acted, power sources process any drain on them (for
handling batteries or other power sources with depletable charge or fuel)
* all entities that died during the course of action processing now get
deleted
Now, this method always places the player last, but allows entities that can
do more than one thing in the time it takes the player to, say, move one
square.
As far as hard numbers go, Pawns (any class of entity that has stats,
usually NPCs and players) have 6 ordinary stats (strength, dexterity,
wisdom, intelligence, constitution, perception) and 12 core skills (hand to
hand, melee, small guns, heavy guns, physics (knowledge of), electronics
(knowledge of and experience in), piloting (ability to), barter (ability
to), speech (effectiveness in communicating), minerology (knowledge of and
ability to extract from ore), metallurgy (knowledge of, ability to extract
from ore, and ability to work with), and manufacturing (knowledge of methods
and ability in manufacturing of materials). Stats range from 1 to 20 and
skills range from 1 to 100 (with no possibility of even temporarily boosting
beyond the max).
A pawn's movement speed is 100 / Dexterity.
A pawn's ability to hack an electronic lock is : Lock.Complexity <=
Electronics.Level to trigger the emergency override, but Lock::Complexity <=
Electronics.Level + Pawn.Intelligence + Pawn.Wisdom if the pawn wants to
rewrite the keyID the lock is expecting (so that hackers can make a lock
accept a keycard with a known keyID).
--
- Pfhoenix
http://pfhoenix.com/adeo
Larry Smith
> In article <Vbkcd.17514$54.2...@typhoon.sonic.net>,
> Ray Dillinger <be...@sonic.net> wrote:
> // at the rate of the fastest player isn't much better;
> // in that case players miss actions because they didn't
> // hit the buttons as fast as machine-gun mike, and the
> // monsters all move at mike's speed, and if anybody
> // pauses for breath, he gets creamed.
>
> There is a sixth option that can lessen machine-gun Mike's advantage...
> namely, allow N actions per time out. With this, MGM has his advantage
> capped, while not feeling constantly hindered by long timeouts for every
> action... and slow players can relax a bit knowing that the game isn't
> going to race totally out of control on them.
Or, one could simply reduce the lethality of the game, so slower game
play doesn't mean automatic death. Of course, that would mean replacing
endless monsters with something resembling an interesting game and that
is definitely not easy. It would require that monsters have some per-
sonality, and agendas of their own.
T(alk ...to: B(alrog: Yo! Listen, we don't need to be enemies!
Balrog: Huh?
Player: Yeah. We worship the same god. We want the same thing...
Balrog: Blood and slaughter?
Player: Exactly!
Balrog: Slaughter you?
Player: No, no, no...OTHER people's blood...
Balrog: What other people?
...and so on.
I vividly recall an encounter in (manual) D&D wherein a fight
with a generic skeleton that turned into a chance for one player
to fast-talk the skeleton into a state of dazed and confused...
--
.-. .-. .---. .---. .-..-.|Experts in Linux/Unix: www.WildOpenSource.com
| |__ / | \| |-< | |-< > / |"Making the bazaar more commonplace"
`----'`-^-'`-'`-'`-'`-' `-' |Check out my new novel: "Cloud Realm" at:
home:www.smith-house.org:8000|http://www.smith-house.org:8000/books/list.html
Brent Ross
// > Ray Dillinger <be...@sonic.net> wrote:
// > // at the rate of the fastest player isn't much better;
// > // in that case players miss actions because they didn't
// > // hit the buttons as fast as machine-gun mike, and the
// > // monsters all move at mike's speed, and if anybody
// > // pauses for breath, he gets creamed.
// >
// > There is a sixth option that can lessen machine-gun Mike's advantage...
// > namely, allow N actions per time out. With this, MGM has his advantage
// > capped, while not feeling constantly hindered by long timeouts for every
// > action... and slow players can relax a bit knowing that the game isn't
// > going to race totally out of control on them.
//
// Or, one could simply reduce the lethality of the game, so slower game
// play doesn't mean automatic death. Of course, that would mean replacing
// endless monsters with something resembling an interesting game and that
// is definitely not easy. It would require that monsters have some per-
// sonality, and agendas of their own.
[snip roleplaying]
The problem is that that's a substantial step forward in AI. Games that
have tried such in the past tend to be very abusable by the players (eg
upset a shopkeeper, jump through hoops #1 through #3 and everything's
fine again, jump through #4 and #5 and she'll be asking to have
your children). So you get munchkins twinking like no tomorrow, and
other players get so familiar with the hoops that they eventually just
pattern match and tokenize entire interactions and zip through them.
RLs get a lot of strength from the fact that their simple abstractions
and a few variables can give quite a complex little simulation with
surprising depth... however, when you try to add an abstraction for
actual intelligent interaction, then that all falls apart.
Basically, in order for such roleplaying to work you need real people
behind everything. Basically, something like one of the less twinky, pure
roleplaying MUDs (ie MUSHs and MOOs more so than LPMuds and Dikus), where
players sit around creating things mostly on the fly for roleplaying's
and entertainment's sake. It just doesn't work well within typical RLs,
so you might as well completely abandon forcing the idea to be a RL and
design freely (but perhaps borrow from RLs later).
Brent Ross
Jeremy Vight
> waits for user input, with a fine granularity (more or less 1 second)
> that makes timing matter: while you decide to talk with a character he
> could have moved away.
> I set aside 3059 because I want time to think, especially during my
> first experience with a new game when I am necessarily slow because I
> have to learn everything.
> (3059 is an otherwise refined, original and interesting game; I just
> dislike the choice of real time interaction.)
> A multiplayer 3059-like game would be even worse: players would try to
> be faster than other players to get an advantage, not only acceptably
> fast for what they are doing.
> With short intervals for actions, shooting and dodging "twitch" action
> is only a few unwise rules away.
>
> Lorenzo Gatti
> ga...@dsdata.it
I have seen too many turn-based RPGs. I really like the idea that you
need to think quickly when you are in a fight and BOTH be
well-prepared in advance... just adds that extra factor you need to
get 'good' at. It also just seems wierd to me to be able to sit down
and go over your options in the middle of a fight... so I made 3059 as
it is :)
Oh yeah, don't forget 3059 offers a 'slow-mo' option you can enable
whenever you want to slow down the game.
And whats wrong with another player being better because he is faster?
Unreal Tournament's whole gameplay relies on how fast you are... I
made 3059 because I didn't want it to be a traditional experience, I
want it to be unique.
I can't please everyone, but I want my game to be as enjoyable as
possible... I don't mind you not choosing to play my game, I just
wanted to defend how I chose to make it.
Thanks, and enjoy! :)
3059's Developer,
Jeremy Vight
Konijn
Or
- Normal roguelike time when a player is 'alone' ( With all the
dificulty of deciding when a player is alone )
- When multiple players are together, the game goes at the speed of
Mighty Mike or 'regular' speed whichever is faster. Monsters AI will
decide that Mighty Mike makes the most moves and is thus most
dangerous and will all charge him. Mighty Mike might still play fast,
but only if he is sure that he can handle the attention.
Cheers,
T.
>
> Brent Ross
David Damerell
>- When multiple players are together, the game goes at the speed of
>Mighty Mike or 'regular' speed whichever is faster. Monsters AI will
>decide that Mighty Mike makes the most moves and is thus most
>dangerous and will all charge him. Mighty Mike might still play fast,
>but only if he is sure that he can handle the attention.
Obvious abuse here; once a player gets increased speed, he just runs
circularly from any number of monsters while his allies, with slower
keystrokes, beat on them.
--
David Damerell <dame...@chiark.greenend.org.uk> Distortion Field!
Larry Smith
> In article <bqGdnROybKu...@adelphia.com>,
> Larry Smith <la...@wildopensource.com> wrote:
> // Or, one could simply reduce the lethality of the game, so slower game
> // play doesn't mean automatic death. Of course, that would mean replacing
> // endless monsters with something resembling an interesting game and that
> // is definitely not easy. It would require that monsters have some per-
> // sonality, and agendas of their own.
> [snip roleplaying]
>
> The problem is that that's a substantial step forward in AI. Games that
> have tried such in the past tend to be very abusable by the players (eg
> upset a shopkeeper, jump through hoops #1 through #3 and everything's
> fine again, jump through #4 and #5 and she'll be asking to have
> your children). So you get munchkins twinking like no tomorrow, and
> other players get so familiar with the hoops that they eventually just
> pattern match and tokenize entire interactions and zip through them.
> RLs get a lot of strength from the fact that their simple abstractions
> and a few variables can give quite a complex little simulation with
> surprising depth... however, when you try to add an abstraction for
> actual intelligent interaction, then that all falls apart.
I think you are being defeatist. Granted it's a non-trivial
problem - if it was easy, every game would already have it.
But I think it is pretty far from insoluable.
The AI problem you allude to above is an artifact of using
a teeny little decision tree. Okay - stick with a small
tree...but make them inheritable. Now a character gets a
behavior pattern based on the character (1st level), the
type (2nd level) the area (3rd level) the planet/continent
(4th level) - plus potential for mixins. Even if each tree
is very small, the combination will lead to emergent behavior
that can be quite unpredictable.
For example, the "Bob" pattern has the character wake up
in the morning, go downstairs, and execute the "shopkeeper"
pattern. The "shopkeeper" pattern sells things, passing
unhandled events back to the "Bob" pattern.
Suppose "Bob" is a retired guardsman. "shopkeeper" punts
to "Bob", and "Bob" punts to "guard". Attacking "Bob"
the shopkeep makes him run to the battleaxe hidden next
to the cashbox. Attacking "Bill" the bartender makes him
run next door to "Bob" screaming for help...
Suppose "Bob" is also an amateur actor. When a certain
clock chime sounds, "Bob" executes the "actor" pattern,
leaves the shop and goes to the theater where he walks
back and forth across a set reciting the part of "Romeo".
Now give "Bob" some other standard patterns - "shopper",
for example - which might be triggered passing a stall
selling apples on his way to the theater - so "Bob" shows
up munching an apple. Give him "generic townsman" and he
can then join mobs (the kind with torches and pitchforks,
not the kind with scripts) and chase miscreants through
the streets.
Individually, each pattern is very simple - but each links
to others via unhandled events or in a hierarchy.
> Basically, in order for such roleplaying to work you need real people
> behind everything.
"Real people" do not want to be shopkeepers in an rpg.
That was Ultima Online's mistake. When you are a shopkeeper,
or close to it, in real life, paying to play one in a game is
not an attractive proposition. Real people want to be the
hero. Which means if you WANT shopkeepers, you had better
figure out some algorithmic way to make them believable.
Brent Ross
Konijn <kon...@gmail.com> wrote:
// bwr...@csclub.uwaterloo.ca (Brent Ross) wrote in message
// news:<cksm5o$ml6$1...@rumours.uwaterloo.ca>...// > Ray Dillinger <be...@sonic.net> wrote:
// > // at the rate of the fastest player isn't much better;
// > // in that case players miss actions because they didn't
// > // hit the buttons as fast as machine-gun mike, and the
// > // monsters all move at mike's speed, and if anybody
// > // pauses for breath, he gets creamed.
// >
// > There is a sixth option that can lessen machine-gun Mike's advantage...
// > namely, allow N actions per time out. With this, MGM has his advantage
// > capped, while not feeling constantly hindered by long timeouts for every
// > action... and slow players can relax a bit knowing that the game isn't
// > going to race totally out of control on them. It's a bit hard to tune,
// > and really is a compromise for both sides (slow players need to rush a
// > bit, fast players experience some jerkiness)... but it is an option.
//
// Or
// - Normal roguelike time when a player is 'alone' ( With all the
// dificulty of deciding when a player is alone )
//
// - When multiple players are together, the game goes at the speed of
// Mighty Mike or 'regular' speed whichever is faster. Monsters AI will
// decide that Mighty Mike makes the most moves and is thus most
// dangerous and will all charge him. Mighty Mike might still play fast,
// but only if he is sure that he can handle the attention.
This is Surreal Time (which was option 5). You're even using a slight
variant of the later suggested small window... instead of having monsters
nearer to the fast player act in a faster time frame, you're having all
the monsters act fast but weight their attention against him. It gives
pretty much the same player, and it certainly has the same abuses... as
pointed out, it probably leads to all kinds of tag play: one player plays
light and quiet and thus is essentially safe, the other plays fast and
draws the attention... <gratuitous monster abuse>... switch roles, repeat.
In short, speed is the most powerful thing in a RL (well known and old
RL proverb... "speed is life"), allowing the players to control monster
speed easily and for free is a big no-no in design. You're simply
better off going real time than doing that, and perhaps writing a Seiken
Densetsu[1]-like instead (although the auto-pause when other players bring
up their spell ring is annoying to player's playing fighter types, IMHO).
Brent Ross
[1] aka Secret of Mana
The Sheep
> Brent Ross wrote:
> Now give "Bob" some other standard patterns - "shopper",
> for example - which might be triggered passing a stall
> selling apples on his way to the theater - so "Bob" shows
> up munching an apple. Give him "generic townsman" and he
> can then join mobs (the kind with torches and pitchforks,
> not the kind with scripts) and chase miscreants through
> the streets.
It's all nice and tidy, but how the `Bob' layer is supposed
to know that `passing a stall' event is in any way different
form `just walking' event, and thus should be cheked against
the `supported/unsupported' list and passed to higher (or
lower, depedning on where your tree's root is) layers of AI?
You'd need a `flat' list of all distinguishable events, and
it's going to be really big if you want all that variety that
benefits from using such a tree.
There's also a problem of conflicts, but you already gave a
partial solution by priorities implied by the layers.
--
Radomir `The Sheep' Dopieralski
You only need pen and paper to do philosophy.
To do mathematics you need pen, paper and a trash can.
Brent Ross
// > // play doesn't mean automatic death. Of course, that would mean replacing
// > // endless monsters with something resembling an interesting game and that
// > // is definitely not easy. It would require that monsters have some per-
// > // sonality, and agendas of their own.
// > [snip roleplaying]
// >
// > The problem is that that's a substantial step forward in AI. Games that
// > have tried such in the past tend to be very abusable by the players (eg
// > upset a shopkeeper, jump through hoops #1 through #3 and everything's
// > fine again, jump through #4 and #5 and she'll be asking to have
// > your children). So you get munchkins twinking like no tomorrow, and
// > other players get so familiar with the hoops that they eventually just
// > pattern match and tokenize entire interactions and zip through them.
// > RLs get a lot of strength from the fact that their simple abstractions
// > and a few variables can give quite a complex little simulation with
// > surprising depth... however, when you try to add an abstraction for
// > actual intelligent interaction, then that all falls apart.
//
// I think you are being defeatist. Granted it's a non-trivial
// problem - if it was easy, every game would already have it.
// But I think it is pretty far from insoluable.
//
// The AI problem you allude to above is an artifact of using
// a teeny little decision tree. Okay - stick with a small
// tree...but make them inheritable. Now a character gets a
// behavior pattern based on the character (1st level), the
// type (2nd level) the area (3rd level) the planet/continent
// (4th level) - plus potential for mixins. Even if each tree
// is very small, the combination will lead to emergent behavior
// that can be quite unpredictable.
You'd like to think so... Go (the board game) has a lot of emergent
behaviour in the games played, however computer opponents play like
rote beginners in the second or third game against a human opponent.
The human brain is an emmensely powerful pattern matcher, in order to
keep it entertained and surprised with social interaction in a game
you need continually changing and adapting interaction. That's quite
simply best done by other human beings... to try and do it with a program
would take years of continuous work, and you'd still end up with human
brains tokenizing the interaction within months (ie "Oh, I though this
was going along pattern A, but it looks like that can branch into the
middle of pattern K"). The lack of abstraction that happens from using
text messages just accelerates the rate (and varying the messages won't
help unless you're willing to put in the years of research and
development into a functional AI storyteller... something, which you
could easily get a PhD for).
Shortly put, it's not worth the large amount of time spent to get
something which will probably be badly abused by the munchkins and
won't have add any replayablity [1]
for anyone else. Replayability is very
important in a RL, it's one of the initial design requirements.
[Bob, the townie snipped ]
That's fine and easy for quaint little features like a townie. There
are easier ways to do the same... it's still not really interaction,
just some things that happen. You can do a lot of quaint things,
but if that's all you do you never actually get a game done. Quaint
stuff adds flavour, but it's only really good the first or second time
(unless it's infrequent enough that the player forgot)... after that
it's not so exciting anymore, and eventually it's filtered out
background noise. In order for quaint stuff to work you need to
have it being added continuously... and a good way to do that is
to farm as much of that work as possible out to the players.
Which is what I meant here:
// > Basically, in order for such roleplaying to work you need real people
// > behind everything
Anyways, my problem was with non-quaint game things like the Balrog
conversion. That's the the sort of thing that ends up as, "Yes, free
pet Balrog" after the first line because the player will recognize the
hoops[2]. In the end I think the most intelligent and sensible thing
for any monster to assume in a social situation is that the player is
trying to trick them and take advantage of the situation by immediately
stabbing them while they're wasting turns talking (the player is far
more cunning and dangerous than any other monster, the world will almost
certainly be a better and safer place[3] with the PC dead).
// "Real people" do not want to be shopkeepers in an rpg.
// That was Ultima Online's mistake. When you are a shopkeeper,
// or close to it, in real life, paying to play one in a game is
// not an attractive proposition. Real people want to be the
// hero. Which means if you WANT shopkeepers, you had better
// figure out some algorithmic way to make them believable.
You could simply make them surly, cynical automatons that are guaranteed
to function correctly without upseting the game. That's the typical way
they were done in MUDs where selling items was important. Making them
more interesting than that was something some later wizard might do while
bored or breaking in their coding chops. In the real roleplaying ones,
selling stuff and being a shopkeeper could be interesting things because
it wasn't a part of a game (when players can create anything they want on
a whim, selling things for gold or buying things isn't inherently a way
to advance... playing a shop scene at that point has to be entertaining,
otherwise you'd just make the room be somewhere else).
Brent Ross
[1] In fact it will probably detract from it if it's central to the game.
[2] Worse yet, a lot of players will probably be spoilt by the "How to
tame a Balrog" spoiler, and you'll end up having to balance the game
against Munchkins with Balrog hordes once this ends up getting posted
as How to Win the Game (tm).
[3] Well, maybe not... but given the typical PC behaviour it's a pretty
good bet and not worth the risk. This of course, assumes that they let
the player near themselves... the most intelligent thing the monsters can
do is grab all the food and good items and run deeper into the dungeon
shouting "The PC is coming". This way the PC either dies of starvation,
goes away, or eventually runs into a huge horde of highly hazardous
monsters before they've had a chance to develop. It's best to nip these
things in the bud while they're still having trouble fighting newts.
Larry Smith
> Dnia Tue, 19 Oct 2004 15:13:20 -0400, Larry Smith napisal(a):
>
>>Brent Ross wrote:
>>Now give "Bob" some other standard patterns - "shopper",
>>for example - which might be triggered passing a stall
>>selling apples on his way to the theater - so "Bob" shows
>>up munching an apple. Give him "generic townsman" and he
>>can then join mobs (the kind with torches and pitchforks,
>>not the kind with scripts) and chase miscreants through
>>the streets.
>
>
> It's all nice and tidy, but how the `Bob' layer is supposed
> to know that `passing a stall' event is in any way different
> form `just walking' event, and thus should be cheked against
> the `supported/unsupported' list and passed to higher (or
> lower, depedning on where your tree's root is) layers of AI?
What, you want a complete implementation plan? I charge
$250 an hour for that. =)
There are any number of ways to do the job. First one to
mind is to trigger a "sees <whatever>" event whenever the
NPC's view is refreshed with something in sight that wasn't
before. Other possibilities include invisible trigger objects
that send events when stepped on. Or maybe the NPC can scan
at certain points to refresh its current view and then trigger
on objects in scope. There are lots of ways.
> You'd need a `flat' list of all distinguishable events, and
> it's going to be really big if you want all that variety that
> benefits from using such a tree.
I don't see that. Unhandled events can just be ignored.
> There's also a problem of conflicts, but you already have a
> partial solution by priorities implied by the layers.
Priorities is one way. Another way might be to allow more
than one pattern to be triggered - whether this is possible
or not might depend on the pattern in force at the time of
the event.
There was a Mac game released many ages ago (for the original
"thin Mac" IIRC) that involved programming little robots using
a visual programming language wherein control flow moved along
arrows from icon to icon on a flat "circuit board", each icon
doing some particular job and moving the flow of control on to
the next icon. Icons could fork or join multiple threads,
transfer control to other boards, and the like. It would not
be hard to imagine something similar for programming these
patterns, with icon/operators for thing like plugging and
unplugging boards implementing certain behaviors, activating
or deactivating boards, scanning for objects, and so on.
Larry Smith
> In article <DsCdnafaHvC...@adelphia.com>,
> Larry Smith <la...@wildopensource.com> wrote:
> You'd like to think so... Go (the board game) has a lot of emergent
> behaviour in the games played, however computer opponents play like
> rote beginners in the second or third game against a human opponent.
> The human brain is an emmensely powerful pattern matcher, in order to
> keep it entertained and surprised with social interaction in a game
> you need continually changing and adapting interaction. That's quite
> simply best done by other human beings...
You know, I recall someone saying this to me about D&D 'round about
1978 or so. "Could never be done by a computer" he said. "You'd
need a million KB." That sounded so daunting when no one could
afford an Apple with a full 48K.
> to try and do it with a program
> would take years of continuous work,
How many years has it taken to get roguelikes to where they
are now? I never said it was trivial. I say it's possible.
> and you'd still end up with human
> brains tokenizing the interaction within months (ie "Oh, I though this
> was going along pattern A, but it looks like that can branch into the
> middle of pattern K").
Anyone can analyze a game to that degree. Roguelikes use
similar algorithms today to generate dungeon levels - does
that mean some people can tell everything they need to know
about a level from looking at the first room? Well, maybe.
But it doesn't change the enjoyment of the game the rest of
us get.
> The lack of abstraction that happens from using
> text messages just accelerates the rate (and varying the messages won't
> help unless you're willing to put in the years of research and
> development into a functional AI storyteller... something, which you
> could easily get a PhD for).
And have. And some of them aren't bad at all. The biggest
problem they have isn't in generating the plot or telling the
story but in parsing out a world description into a form that
can make sense to the AI. Roguelikes have that already.
> Shortly put, it's not worth the large amount of time spent to get
> something which will probably be badly abused by the munchkins and
> won't have add any replayablity [1]
> for anyone else. Replayability is very
> important in a RL, it's one of the initial design requirements.
I see no reason to assume this, either. Firstly, properly done
it is no more susceptible to munchkins than the current algorithms
are within their perview. Secondly, we aren't limited to just one
pattern in each class. Lastly, the complexity comes from the
interactions in certain behavior patterns. And remember these
_are_ just patterns. The actual bahavior that results is influenced
by the current situation, past history, even things like current
mood and status.
> [Bob, the townie snipped ]
> That's fine and easy for quaint little features like a townie. There
> are easier ways to do the same... it's still not really interaction,
> just some things that happen. You can do a lot of quaint things,
> but if that's all you do you never actually get a game done. Quaint
> stuff adds flavour, but it's only really good the first or second time
> (unless it's infrequent enough that the player forgot)... after that
> it's not so exciting anymore, and eventually it's filtered out
> background noise.
"Quaint." Roguelikes are already complex enough to have considerable
emergent behavior. Adding this mechanism will increase that by an
order of magnitude or more.
This is a _general_ mechanism, not just something for steering NPC's
around in a fixed area. It can be used for physical behavior, con-
versation, and even for plots and area layout. Right now, RLs can't
easily control what features might exist in a given area. Imagine
a RL of Chalker's "Well of Souls" books. Here the very rules of the
universe change from zone to zone. In the above system, this is
easy if the features are each zone are defined in patterns used to
generate them.
Granted it's a new way to think about this stuff. Well, so was
object oriented programming back when structured programming was
the standard.
> Anyways, my problem was with non-quaint game things like the Balrog
> conversion. That's the the sort of thing that ends up as, "Yes, free
> pet Balrog" after the first line because the player will recognize the
> hoops[2].
It might be a free pet Balrog. Or it might eat you if you're
wrong. Or drag you off to the nethermost region of hell (which
would have it's own pattern for generation, of course). It
could also be that the Balrog might be thinking "Yes, free pet
human..."
You are assuming you know, or can guess, all the patterns that
might exist in a given situation. That would be a poor implementation.
Done well, you could never know for sure what patterns are in force.
Some may only come into force once in a blue moon, others may be so
similar to other patterns they can't be easily distinguished until
they interact with yet more patterns - and by then it may be too late.
> In the end I think the most intelligent and sensible thing
> for any monster to assume in a social situation is that the player is
> trying to trick them and take advantage of the situation by immediately
> stabbing them while they're wasting turns talking (the player is far
> more cunning and dangerous than any other monster, the world will almost
> certainly be a better and safer place[3] with the PC dead).
Do you want to hack-n-slash or role play? If the former, I can
see why you see little use for what I suggest, but then this whole
discussion is pointless, we diverged at the very first statement
"make the game less lethal".
> // "Real people" do not want to be shopkeepers in an rpg.
> // That was Ultima Online's mistake. When you are a shopkeeper,
> // or close to it, in real life, paying to play one in a game is
> // not an attractive proposition. Real people want to be the
> // hero. Which means if you WANT shopkeepers, you had better
> // figure out some algorithmic way to make them believable.
>
> You could simply make them surly, cynical automatons that are guaranteed
> to function correctly without upseting the game.
And so we once again turn away from role playing and back to
simple hack-n-slash. If that satisfies you, fine, I'm not
putting a gun to your head and demanding you implement the
idea.
> [2] Worse yet, a lot of players will probably be spoilt by the "How to
> tame a Balrog" spoiler, and you'll end up having to balance the game
> against Munchkins with Balrog hordes once this ends up getting posted
> as How to Win the Game (tm).
Any method of taming a Balrog should not work twice and
never easily translate to another game.
Brent Ross
// > behaviour in the games played, however computer opponents play like
// > rote beginners in the second or third game against a human opponent.
// > The human brain is an emmensely powerful pattern matcher, in order to
// > keep it entertained and surprised with social interaction in a game
// > you need continually changing and adapting interaction. That's quite
// > simply best done by other human beings...
//
// You know, I recall someone saying this to me about D&D 'round about
// 1978 or so. "Could never be done by a computer" he said. "You'd
// need a million KB." That sounded so daunting when no one could
// afford an Apple with a full 48K.
D&D still isn't doable... you can grind the mechanics, but the computer
can't run let alone play the game. That might be doable before Go,
nd that's not going to happen for a long, long time (several decades
minimum). Even when it is done, it's not going to be trivial... far
from something that's a casual project.
// > to try and do it with a program
// > would take years of continuous work,
//
// How many years has it taken to get roguelikes to where they
// are now? I never said it was trivial. I say it's possible.
Years since the start of a genre are radically differnent from man years
on a casual project (especially when you need doctorate level stuff
to really do the job... most people would rather take that and get the
degree and turn it into something practical and worth serious money).
// > and you'd still end up with human
// > brains tokenizing the interaction within months (ie "Oh, I though this
// > was going along pattern A, but it looks like that can branch into the
// > middle of pattern K").
//
// Anyone can analyze a game to that degree. Roguelikes use
// similar algorithms today to generate dungeon levels - does
// that mean some people can tell everything they need to know
// about a level from looking at the first room? Well, maybe.
// But it doesn't change the enjoyment of the game the rest of
// us get.
The abstraction of level creation is very different from trying to create
social interaction. A few variables and a new routine and you've created
a zillion new permutations... spend an hour every now and then to do
this and the complexity of the dungeons keeps continually growing.
Sure the player will see patterns in them and sort them into a much
smaller set, but a single pillar left in a room can give a radically
different situation when played. Scoial interaction code just can't give
you the same bag for the buck unless you've got an AI story generator
and Turing Test success to back it up... the returns just die off too
quickly if you try and do it yourself.
// > The lack of abstraction that happens from using
// > text messages just accelerates the rate (and varying the messages won't
// > help unless you're willing to put in the years of research and
// > development into a functional AI storyteller... something, which you
// > could easily get a PhD for).
//
// And have. And some of them aren't bad at all. The biggest
// problem they have isn't in generating the plot or telling the
// story but in parsing out a world description into a form that
// can make sense to the AI. Roguelikes have that already.
They don't have what I'm talking about... they're so unbelievably far
from it in fact (of the dozens I've played). What I'm talking about
is the ability of the game to actually play D&D. To be able to take
something it never expected and outside the box and act like a real
human DM and work it in. If any game does have anything close to that,
I'd suggest the author run out and get a patent and start raking in
the billions... the applications for such technology go far beyond
something as trivial as a game (it's almost a waste really).
// > Shortly put, it's not worth the large amount of time spent to get
// > something which will probably be badly abused by the munchkins and
// > won't have add any replayablity [1]
// > for anyone else. Replayability is very
// > important in a RL, it's one of the initial design requirements.
//
// I see no reason to assume this, either. Firstly, properly done
// it is no more susceptible to munchkins than the current algorithms
// are within their perview.
Typically it's more susceptible. Munchkins don't need to code dive
to break these systems... the complexity of developing them typically
means that they're the simplest and weakest link. It takes
exponentially more time to keep social interaction code up to par.
// Secondly, we aren't limited to just one
// pattern in each class. Lastly, the complexity comes from the
// interactions in certain behavior patterns. And remember these
// _are_ just patterns. The actual bahavior that results is influenced
// by the current situation, past history, even things like current
// mood and status.
Which is a lot of work for something which people will be able to see
right through to the salient features. You're right that this has
been done in games before... and they all ramble on about how smart the
code is, and how it's better than anything before, and how the number
of resultant permutations is ungodly huge. And yet the end result and
appearance to the user is always pretty much the same: abuse and a
couple extra seconds of replayability.
// > [Bob, the townie snipped ]
// > That's fine and easy for quaint little features like a townie. There
// > are easier ways to do the same... it's still not really interaction,
// > just some things that happen. You can do a lot of quaint things,
// > but if that's all you do you never actually get a game done. Quaint
// > stuff adds flavour, but it's only really good the first or second time
// > (unless it's infrequent enough that the player forgot)... after that
// > it's not so exciting anymore, and eventually it's filtered out
// > background noise.
//
// "Quaint."
Bob is a very quaint feature... he's capable of giving a moment
of joy. He's also got a lot of wasted complexity because he'll
never be worth more than that.
// Roguelikes are already complex enough to have considerable
// emergent behavior. Adding this mechanism will increase that by an
// order of magnitude or more.
It will increase the workload by a magnitude for something that's
worth a second. A Bob was ran to his store and grabbed a battle-axe...
there, fun done, no townie will ever surprise me again (nor any
other monster that moves in a similar way). Ho hum... quaint.
// Granted it's a new way to think about this stuff. Well, so was
// object oriented programming back when structured programming was
// the standard.
New? This idea was old years ago when this newsgroup started. Every
few monthes someone posts this idea of having a few motivation
variables and patterns designed to give emergent natural behaviours.
It's quaint... it will look nice... it adds a little reality to the
simulation. Fine and dandy.... but you're no where near what you
need for achieveing your solution to the original problem. This
isn't going to shift complexity away and make multiplayer roguelikes
more easily playable.
// > Anyways, my problem was with non-quaint game things like the Balrog
// > conversion. That's the the sort of thing that ends up as, "Yes, free
// > pet Balrog" after the first line because the player will recognize the
// > hoops[2].
//
// It might be a free pet Balrog. Or it might eat you if you're
// wrong. Or drag you off to the nethermost region of hell (which
// would have it's own pattern for generation, of course). It
// could also be that the Balrog might be thinking "Yes, free pet
// human..."
You missed the point. If the end result is random like that then the
player will almost certainly recognize that almost as fast. The problem
is that if there is a way to navigate these interactions in a near
reliable way... the players will catch on and abuse such a system
immediately. End result: a random system the players mostly ignore
(I could talk to the Balrog, but it's pretty much meaningless if not
downright dangerous and potentially stupid), or the player learns the
hoops and zips through conversations so fast you might as well add a
"tame Balrog" command. Of the dozens of games with supposedly complex
social interaction and hyperadvanced AIs I've played... they've all pretty
much one or the other (or a bit of both, which is often simpler yet).
Software technology is no where near being otherwise.
// You are assuming you know, or can guess, all the patterns that
// might exist in a given situation. That would be a poor implementation.
If you have a game where all *relevant* pattern features weren't obvious,
the program would be remarkably intelligent... see patent and billions
$$$. Anything which can be easily written, with a couple variables
and functions and emergent features can't apply. The simplicity of
the design will prevent it from being able to do that... people will
see right through the surface to the core (because they are intelligent
pattern matching machines... the exact things that modern computers and
software are not).
// > In the end I think the most intelligent and sensible thing
// > for any monster to assume in a social situation is that the player is
// > trying to trick them and take advantage of the situation by immediately
// > stabbing them while they're wasting turns talking (the player is far
// > more cunning and dangerous than any other monster, the world will almost
// > certainly be a better and safer place[3] with the PC dead).
//
// Do you want to hack-n-slash or role play?
Role play... especially if there's supposted to be social interaction.
// If the former, I can
// see why you see little use for what I suggest, but then this whole
// discussion is pointless, we diverged at the very first statement
// "make the game less lethal".
It was after that when you jumped into replacing some original game
complexity with social complexity. Making the game less lethal
by making it a lot simpler is something I'd alluded to before.
Simplification is pretty much a given if you're going with a timeout
system (because the number of actions taken in a RL is large, and
current designs have some much more complicated than others to
perform... running down the long hallways needs to be quick, but
there needs to be time for spellcasters to cast spells, or players
to deal with their inventory). Reducing lethality is one easy
way to increase simplicity (a safer game allows for more leeway
with player error and slowness).
// > // "Real people" do not want to be shopkeepers in an rpg.
// > // That was Ultima Online's mistake. When you are a shopkeeper,
// > // or close to it, in real life, paying to play one in a game is
// > // not an attractive proposition. Real people want to be the
// > // hero. Which means if you WANT shopkeepers, you had better
// > // figure out some algorithmic way to make them believable.
// >
// > You could simply make them surly, cynical automatons that are guaranteed
// > to function correctly without upseting the game.
//
// And so we once again turn away from role playing and back to
// simple hack-n-slash. If that satisfies you, fine, I'm not
// putting a gun to your head and demanding you implement the
// idea.
Implement forcing real players to play as Shopkeepers? Wouldn't
even think of it.
// > [2] Worse yet, a lot of players will probably be spoilt by the "How to
// > tame a Balrog" spoiler, and you'll end up having to balance the game
// > against Munchkins with Balrog hordes once this ends up getting posted
// > as How to Win the Game (tm).
//
// Any method of taming a Balrog should not work twice and
// never easily translate to another game.
So it's a random thing that player's will probably never do, since one
can assume that most Balrog encounters end up in someone's death and it's
best to jump to the old standard hack-n-slash rather than let the Balrog
go there first. Pattern solved, never have to think about it again.
Brent Ross
The Sheep
> The Sheep wrote:
>> Dnia Tue, 19 Oct 2004 15:13:20 -0400, Larry Smith napisal(a):
<snip>
>> You'd need a `flat' list of all distinguishable events, and
>> it's going to be really big if you want all that variety that
>> benefits from using such a tree.
>
> I don't see that. Unhandled events can just be ignored.
Yes, but you have more specific and less specific events.
The AI must know what events it's interested in -- and it
depends on the leafs of that tree you're talking about.
Another way would be to generate a thousand different every
turn, and ignore 99% of them. But that's when you need that
2Ghz of your processor ^^))
Konijn
> Konijn <kon...@gmail.com> wrote:
> >- When multiple players are together, the game goes at the speed of
> >Mighty Mike or 'regular' speed whichever is faster. Monsters AI will
> >decide that Mighty Mike makes the most moves and is thus most
> >dangerous and will all charge him. Mighty Mike might still play fast,
> >but only if he is sure that he can handle the attention.
>
> Obvious abuse here; once a player gets increased speed, he just runs
> circularly from any number of monsters while his allies, with slower
> keystrokes, beat on them.
I only play angband, so I could have a wrong view on it. I have just
tried to circle around a bunch of summoned orcs and it is just
impossible to circle around them, I'm sure someone with math skills
could even prove that.
Note that all monsters ( the game ) will increase speed as mike does.
Also I didnt mean that all other AI functionality should be shut off,
just that it should be an important factor in choosing targets.
Cheers,
T.
Larry Smith
> D&D still isn't doable... you can grind the mechanics, but the computer
> can't run let alone play the game. That might be doable before Go,
> nd that's not going to happen for a long, long time (several decades
> minimum). Even when it is done, it's not going to be trivial... far
> from something that's a casual project.
Depends on your expectations. Actually, quite a few computer games
nowadays can produce a gaming experience the equal of many DMs from
the 70's. Granted the art of GMing has advanced since then, but the
observation remains valid.
> // > to try and do it with a program
> // > would take years of continuous work,
> //
> // How many years has it taken to get roguelikes to where they
> // are now? I never said it was trivial. I say it's possible.
>
> Years since the start of a genre are radically differnent from man years
> on a casual project (especially when you need doctorate level stuff
> to really do the job... most people would rather take that and get the
> degree and turn it into something practical and worth serious money).
Let's not argue. Let's do. Or do not.
I am going to fork this thread and try to generate some ideas.
Let's see what the developers of these games think.
David Damerell
>David Damerell <dame...@chiark.greenend.org.uk> wrote in message:
>>Konijn <kon...@gmail.com> wrote:
>>>- When multiple players are together, the game goes at the speed of
>>>Mighty Mike or 'regular' speed whichever is faster. Monsters AI will
>>>decide that Mighty Mike makes the most moves and is thus most
>>>dangerous and will all charge him. Mighty Mike might still play fast,
>>>but only if he is sure that he can handle the attention.
>>Obvious abuse here; once a player gets increased speed, he just runs
>>circularly from any number of monsters while his allies, with slower
>>keystrokes, beat on them.
>I only play angband, so I could have a wrong view on it. I have just
>tried to circle around a bunch of summoned orcs and it is just
>impossible to circle around them, I'm sure someone with math skills
>could even prove that.
I'm sure they couldn't, since such circling is a common tactic in Crawl
and NetHack's BigRoom. Remember, Mike has increased speed; the monsters
are on the same tempo as him, but still don't move as quickly.
Even failing such a convenient situation, Mike can jiggle back and forth
to drag a string of monsters through his allies, who beat on them with
slightly slower keystrokes. A badly wounded player can stop typing and
wait for some healing. A party with a heavily armoured damage-sink player
can ensure they take the lion's share of punishment without actually
having to have good tactical positioning.
--
David Damerell <dame...@chiark.greenend.org.uk> flcl?
Larry Smith
Brent Ross and I have been arguing the merits of using patterns to control
the behavior of various game aspects. He thinks it will provide little
return for the effort, I disagree. I'd like to start a thread to bang
around a few ideas about an implementation, and see what developeds are
thinking.
Firstly, I envision a basic pattern. This is essentially a program that
is invoked when an event occurs within the perception of character or
feature controlled by this pattern.
The complexity comes in when various patterns are combined into a given
character. There are several ways this could be done. One way is hier-
archical: some patterns are more important than others, when a given
event matches more than one active pattern, the first one in the hierarchy
takes precedence. Another way is to allow patterns to load and unload
other patterns - that is, one response to a given event might result in
adding a new pattern of behavior.
Thus a NPC "brain" is a box consisting of some number of pattern "cards",
each describing a series of responses to various event stimuli. Cards
(patterns) can be added or removed under program control. Cards have a
particular order, so a given event only triggers the earliest matching
pattern of behavior. A pattern that is triggered may either succeed or
fail - if it succeeds, control ends, if it fails, control passes to the
next matching pattern down the line.
Patterns could have wild cards to match more general or more specific
events. A specific event might be "addressedby-charlie". A more
general event might be "addressedby-<party member>". A completely
general event might be "addressedby-<anyone>" - in this example, the
<> terms signify classes.
QUESTION: what kind of events would be needed? "addressed-by" is one
triggered by t(alk. Some others might be "sees" for objects that
appear in the NPC's visual range, "smells", and "hears" likewise.
"isattacked" is another obvious one. What else comes to mind?
Brain boxes are loaded with cards when the NPC is created. The cards
that are loaded depend on the behaviors we wish to program - they could
be location-based (all NPC's in city Xagoomvir get this card), or class
based (all warriors get this card) or specific (only innkeeper Billy
the Pig) or even random by any of the foregoing.
QUESTION: what other kind of classes might one wish to specify in order
to program in this fashion?
If the "NPC" is, in fact, an AI that helps build an area of the world,
patterns can be used to change the way various portions of the world are
built.
QUESTION: what kind of patterns would be useful for such?
Brent Ross
// the behavior of various game aspects. He thinks it will provide little
// return for the effort, I disagree.
You're still not undersatnding my position. My problem isn't with using
patterns to generate behaviours. It has to do with the goal. If the
goal is to add some quaintness and realistic monster behaviour[1] to the
game then you get a lot of bang out of combining a few patterns together.
If, on the other hand, you're trying to generate complexity in social
roleplaying aspects with them you get very diminshed returns on your work.
Those are very different goals. Patterned behaviour is an underlying
thing... it acts pretty much transparently and occasionally subtlty
affects things, but it's not something the player plays directly with as
part of the game. It's a bit of background that adds a little bit to
people who are roleplaying[2]. Roleplaying still is really left to the
player's own head, which is the only way it can work (short of getting
other real people or sentient machines involved).
Social interactions like actually handling conversation with monsters
isn't subtle... it's direct, it's upfront, and if it's an active part
of the game it takes an awful lot of scripting to get the density
to be actually interesting and functional from a roleplaying point.
Munchkins ride the system like any other (only better because it's
inherently easy to see and less dense than most), and Real Roleplayers
see the interaction as trivial, constricting, pointless, or random.
This is unsolvable until you have a program that's as good as a real DM.
In short, if your goals are quaint features and realistic world simulation
and monster behaviour you can easily win with this type of approach.
But you're pretty much doomed if you try to make it more than that and
a serious part of the upfront game.
Brent Ross
[1] I've always been against trying for fully intelligent monster
behaviour... truly intelligent monsters evacuate the level and/or take
advantage of their massive numbers to overwelm the player, and different AIs
just result in what we around here call "Monsters behave stupidly in a
different way" (after the Angband option). I'm an advocate for "monsters
behave appropriately"... orcs should behave like orcs, it's that simple.
[2] ... as opposed to rollplaying (for those who haven't heard that yet)
konijn_
> Konijn wrote:
> >David Damerell wrote in message:
> >>Konijn wrote:
> >>>- When multiple players are together, the game goes at the speed of
> >>>Mighty Mike or 'regular' speed whichever is faster. Monsters AI will
> >>>decide that Mighty Mike makes the most moves and is thus most
> >>>dangerous and will all charge him. Mighty Mike might still play fast,
> >>>but only if he is sure that he can handle the attention.
> >>Obvious abuse here; once a player gets increased speed, he just runs
> >>circularly from any number of monsters while his allies, with slower
> >>keystrokes, beat on them.
> >I only play angband, so I could have a wrong view on it. I have just
> >tried to circle around a bunch of summoned orcs and it is just
> >impossible to circle around them, I'm sure someone with math skills
> >could even prove that.
>
> I'm sure they couldn't, since such circling is a common tactic in Crawl
> and NetHack's BigRoom. Remember, Mike has increased speed; the monsters
> are on the same tempo as him, but still don't move as quickly.
Yeah, no such big rooms in Angband. And most monster move at the exact same
speed of the player or faster.
> Even failing such a convenient situation, Mike can jiggle back and forth
> to drag a string of monsters through his allies, who beat on them with
> slightly slower keystrokes. A badly wounded player can stop typing and
> wait for some healing. A party with a heavily armoured damage-sink player
> can ensure they take the lion's share of punishment without actually
> having to have good tactical positioning.
Yes, in mudding slang that heavily armoured guy is the 'tank'.
If I would ever ( dont think I will ) see players cooperating in such a fashion
to play an online rogue game, I would say that is a success, not a failure.
Also, keep in mind that the playingspeed of the player is only one of the AI
factors to decide target, not the only factor, so monsters should still
consider slow moving almost dead players.
Cheers,
T.
> --
> David Damerell flcl?
>
>
Brent Ross
//
// Yes, in mudding slang that heavily armoured guy is the 'tank'.
// If I would ever ( dont think I will ) see players cooperating in such a fashion
// to play an online rogue game, I would say that is a success, not a failure.
Actually, in this case it's a huge failure. The players are cooperating
not to *play* the game, but to intentionally *break* the game. Having
the players distorting the nature of space time is not a success,
it's a huge problem.
Brent Ross
R. Dan Henry
(Brent Ross) wrote:
>In article <cl6pip$2dv$1...@news.vol.cz>, konijn_ <kende...@hotmail.com> wrote:
>//
>// Yes, in mudding slang that heavily armoured guy is the 'tank'.
>// If I would ever ( dont think I will ) see players cooperating in such a fashion
>// to play an online rogue game, I would say that is a success, not a failure.
>
>Actually, in this case it's a huge failure. The players are cooperating
>not to *play* the game, but to intentionally *break* the game.
Finding the exploits is half the fun of roguelike.
>Having
>the players distorting the nature of space time is not a success,
>it's a huge problem.
It sure is a success, if you expected it! If you get surprised by the
existence of an exploit, then that's a failure to anticipate on your
part, but player's using a game feature as anticipated is never a
failure. It isn't realistic, but if that's your objection, shame on
you. You should know better.
R. Dan Henry
danh...@inreach.com
The Sheep
> Larry Smith <la...@wildopensource.com> wrote:
> [1] I've always been against trying for fully intelligent monster
> behaviour... truly intelligent monsters evacuate the level and/or take
> advantage of their massive numbers to overwelm the player, and different AIs
> just result in what we around here call "Monsters behave stupidly in a
> different way" (after the Angband option). I'm an advocate for "monsters
> behave appropriately"... orcs should behave like orcs, it's that simple.
Why do you always assume that:
1) All the monsters should cooperate? Most of the time tehy are monsters
also to themselves, so they are pretty happy when their neighbour gets
killed.
2) They might try to use the player character for their own purposes --
like aforementioned killing of hated neighbours.
3) They may underestimate the player character.
4) They may other goals than just slaying the player character. Living
creatures tend to defend their homes with their lives. I think it's
more likely that they wait, hoping the player character won't come,
and fight when he comes, instead of abandoning all their possesions and
teritory they fought so hard for.
5) It's a cruel world down in the dungeons, and a few hundred dead
monsters along the path of player character is not going to cause a panic
on lower levels -- it's just a norm.
Brent Ross
R. Dan Henry <danh...@inreach.com> wrote:
// (Brent Ross) wrote:
// >Having
// >the players distorting the nature of space time is not a success,
// >it's a huge problem.
//
// It sure is a success, if you expected it!
Only if we're doing sci-fi time stuff[1]. If the characters aren't suppost
to have power over time and space and a design essentially gives that
to them for free it's a huge failure and should be scrapped before it
sees the light of day. Time is too big in RLs to let such an abuse get
off the drawing board.
// If you get surprised by the
// existence of an exploit, then that's a failure to anticipate on your
// part, but player's using a game feature as anticipated is never a
// failure.
But we're not talking about that... in this case we're talking about
the players cooperating[2] to abuse the system as a reason for why it
was flawed. The fact that they can easily do such a thing is a sign
that the designer needs to go back to the drawing board... they've
only created a horrid munchkin metagame that really isn't that fun.
// It isn't realistic, but if that's your objection, shame on
// you. You should know better.
My objection is to the idea that group abuse equals success.
Realism doesn't have much to do with that (the game is plain
broken)... although the game also has realism stretched beyond
any reasonable bound (some abstraction is reasonable, but some
sense of realism[3] is always a requirement).
Brent Ross
[1] Which could be fun, but if the players have the ability to freely
adjust time on their whim like in this thread, it wouldn't be.
[2] Perhaps collusion would be a better word here, because people
have this automatic "cooperation is always good" moral drilled into
their head from childhood. In this case, the players are working
together to break the game under the designers nose.
[3] Realism to the game world. The (meta)physics should be
understandable, known to the designer, and consistant (even if the
consistancy is defined as "Dadarogue" it's still consistant to
the goal of Dada).
Brent Ross
The Sheep <sh...@sheep.prv.pl> wrote:
// Dnia Wed, 20 Oct 2004 19:25:07 +0000 (UTC), Brent Ross napisal(a):
// > Larry Smith <la...@wildopensource.com> wrote:
// > [1] I've always been against trying for fully intelligent monster
// > behaviour... truly intelligent monsters evacuate the level and/or take
// > advantage of their massive numbers to overwelm the player, and different AIs
// > just result in what we around here call "Monsters behave stupidly in a
// > different way" (after the Angband option). I'm an advocate for "monsters
// > behave appropriately"... orcs should behave like orcs, it's that simple.
//
// Why do you always assume that:
// 1) All the monsters should cooperate?
Because the monsters are "perfectly intelligent" like the players (and
some designers) ask for. As such, they realize that the most dangerous
thing to their regular life is the PC... this monster will exponentially
grow in power if it's allowed to kill any of us! So even if Joe Kobold
thinks he has a chance against him, being perfectly intelligent he
realizes that the percentage is much higher if he can lure the threat
deeper into the dungeon where bigger monsters might have a sure kill (if
Joe dies by the PCs hand, his blood might give the PC enough power that
no Kobold will survive it's wrath!). Thus the monsters call a truce
(being perfectly intelligent this is easier for them than real world
politicians) until the menance is dealt with. Once the PC is dead they
can go back to hating each other in a civilized, less apocalyptic fashion.
// Most of the time tehy are monsters
// also to themselves, so they are pretty happy when their neighbour gets
// killed.
That's appropriate behaviour, not intelligent.
// 2) They might try to use the player character for their own purposes --
// like aforementioned killing of hated neighbours.
That's appropriate behaviour, not intelligent.
// 3) They may underestimate the player character.
That's appropriate behaviour, not intelligent.
// 4) They may other goals than just slaying the player character. Living
// creatures tend to defend their homes with their lives. I think it's
// more likely that they wait, hoping the player character won't come,
// and fight when he comes, instead of abandoning all their possesions and
// teritory they fought so hard for.
That's appropriate behaviour, not intelligent.
You seemed to have missed the point (maybe you have it now). Monsters
behaving appropriately is not anywhere near the same as intelligently.
The monsters vastly out number the PC, give an actual human control
of them to use intelligently against the PC and the PC dies quick and
horribly.
I'm all for having monsters behave exactly as your points suggest... and
the suggestion that started this subthread can do that nicely without
too much work (which is my point... if his point is to do this for
some appropriate and quaint monster behaviours, I don't disagree like
he implied). Although, there is some diminishing returns that will show
up fairly early because a lot of small factors will wash each other out
in such a system (more is less/less is more)... there's a balance that
needs to be found.
// 5) It's a cruel world down in the dungeons, and a few hundred dead
// monsters along the path of player character is not going to cause a panic
// on lower levels -- it's just a norm.
But the PC gains power from the blood of the slain. Whereas, he used to
susceptible to newts, he's now killing giants (seven with one blow!).
Panic! Panic! Everyone for himself! Run to the four corners of the
world and tell them to never let one of these things hurt a fly!
Brent Ross
The Sheep
It was `intelligent' not `perfectly intelligent' last time I read it, but
it's okay. I agree that not all the monsters in the dungeon should behave
equally smart, but you should have room for leveling.
Some monsters (mostly humanoids, as I see) should be intelligent. But
they should still have their own goals, not necessarily killing the
player. Calling a truce? What for? This guy is already doing things
I wanted to be done for years, and with some little tricks he will spare
me, and I can claim all the empty terriority for myself.
Also, it takes quite some time to observe that the player is growing in
power. Monsters that meet the player character usually don't live long
enough to tell the tale. Sure, `perfectly intelligent' monster will
know, but will `just intelligent' one know too? If so, see few lines later.
> // Most of the time tehy are monsters
> // also to themselves, so they are pretty happy when their neighbour gets
> // killed.
>
> That's appropriate behaviour, not intelligent.
Why isn't that intelligent? Intelligence has nothing to do with malice.
You can be both intelligent and malicious.
> // 2) They might try to use the player character for their own purposes --
> // like aforementioned killing of hated neighbours.
>
> That's appropriate behaviour, not intelligent.
I think it's more intelligent to try to use something at your advantage
instead of getting rid of it right away `cus it might be dangerous'.
> // 3) They may underestimate the player character.
>
> That's appropriate behaviour, not intelligent.
This again has nothng to do with intelligence. It's knowledge, and they
might not have it if there was no way to get it.
> // 4) They may other goals than just slaying the player character. Living
> // creatures tend to defend their homes with their lives. I think it's
> // more likely that they wait, hoping the player character won't come,
> // and fight when he comes, instead of abandoning all their possesions and
> // teritory they fought so hard for.
>
> That's appropriate behaviour, not intelligent.
So you call a mania of slaying `a chosen one' intelligent, but a desire to
enjoy your monster life, do something meaningful, seize control of the
whole dungeon -- non-intelligent?
> You seemed to have missed the point (maybe you have it now). Monsters
> behaving appropriately is not anywhere near the same as intelligently.
Those are two different scales. The monster might behave appropriately
and intelligently, as well as it might behave appropriately and stupidly.
> The monsters vastly out number the PC,
Assuming they unite.
> give an actual human control
> of them to use intelligently against the PC and the PC dies quick and
> horribly.
And the monsters gain nothing. They are back to their old, boring life.
Hardly intelligent.
> I'm all for having monsters behave exactly as your points suggest... and
> the suggestion that started this subthread can do that nicely without
> too much work (which is my point... if his point is to do this for
> some appropriate and quaint monster behaviours, I don't disagree like
> he implied). Although, there is some diminishing returns that will show
> up fairly early because a lot of small factors will wash each other out
> in such a system (more is less/less is more)... there's a balance that
> needs to be found.
True that this can add flovour to monster behavior. And it's good. It's
one thing.
Intelligent behavior is another thing, and it's supposed to make the
monster (note that it's the monster, not the mosnters) a harder opponent.
Offcourse he should still behave appropriately at the same time. And
you don't want all the monsters to be intelligent.
> // 5) It's a cruel world down in the dungeons, and a few hundred dead
> // monsters along the path of player character is not going to cause a panic
> // on lower levels -- it's just a norm.
>
> But the PC gains power from the blood of the slain. Whereas, he used to
> susceptible to newts, he's now killing giants (seven with one blow!).
> Panic! Panic! Everyone for himself! Run to the four corners of the
> world and tell them to never let one of these things hurt a fly!
Yeah, and those monsters at the gates of hell, will mercifully slay
the intruder because we asked them to do it. Doubt it.
There's no cooperation between dungeon lowlife and great archdaemons.
At least I don't believe in such thing. And it's not a question of
intelligence.
How about this scenario:
I'll gladly give him my life if this will increase his chances of slaying
that evil overlord that treats all my race as slaves.
It's pretty intelligent too. Even more, becuase it's not pure
self-preservation instinct.
Brent Ross
// Dnia Thu, 21 Oct 2004 08:25:26 +0000 (UTC), Brent Ross napisal(a):
// > The Sheep <sh...@sheep.prv.pl> wrote:
// > // Dnia Wed, 20 Oct 2004 19:25:07 +0000 (UTC), Brent Ross napisal(a):
// > // > [1] I've always been against trying for fully intelligent monster
// > // > behaviour... truly intelligent monsters evacuate the level and/or take
// > // > advantage of their massive numbers to overwelm the player, and
// different AIs
// > // > just result in what we around here call "Monsters behave stupidly in a
// > // > different way" (after the Angband option). I'm an advocate for "monsters
// > // > behave appropriately"... orcs should behave like orcs, it's that simple.
// It was `intelligent' not `perfectly intelligent' last time I read it, but
// it's okay.
In the footnote I used "fully" and "truly". The thing is that there
are people that want the monsters to play with the intelligence of
a human player. The problem with that is that they outnumber the PC
by thousands to 1 and the end result is that such a game would only
work as a Fung Shui (the RPG) roguelike (in short, almost all monsters
would be mooks and trivially killable... there would be maybe a dozen
odd named monsters who would be a threat).
// Calling a truce? What for? This guy is already doing things
// I wanted to be done for years, and with some little tricks he will spare
// me, and I can claim all the empty terriority for myself.
Because the PC is an apocalyptic element. Not only is he doing things
you want done, he's probably doing things you *don't* want done. If you
let him be, he will become more powerful than you can imagine, and
then we'll all potentially be doomed (because he will be unstoppable).
Since the monsters are intelligent, they are capable of making a truce
for the sake of one day returning to their regularly scheduled genocides
in civilized fashion. The PC is too dangerous and a huge wildard (who
knows what such a powerful monster will do?) to let be... everyone is
better off without him in the world.
// Also, it takes quite some time to observe that the player is growing in
// power. Monsters that meet the player character usually don't live long
// enough to tell the tale. Sure, `perfectly intelligent' monster will
// know, but will `just intelligent' one know too? If so, see few lines later.
Yes... we're talking true intelligent monster behaviour here from the
point of view of AI. They behave like they're being played by sentient
beings, and so they know the basic rules of the game ("PC gains power
through the blood of those it slays") as well as the player does.
// > // Most of the time tehy are monsters
// > // also to themselves, so they are pretty happy when their neighbour gets
// > // killed.
// >
// > That's appropriate behaviour, not intelligent.
// Why isn't that intelligent? Intelligence has nothing to do with malice.
// You can be both intelligent and malicious.
It's not intelligent... it's appropriate. Appropriately intelligent for
a monster from the point of view *inside* the game, not from the point
of view of the game AI. You seem to have completely missed that that's
what we're talking about. Apporopriate behaviour is different from
intelligent behaviour because the monsters are acting naturally in their
own world... sometimes that can be intelligent behaviour from the game's
point of view as well, but in this case it's not (the monsters are
far better off combining forces and ganging up on the player... it's
only that that would be inappropriate for them to do in character that
prevents that).
// > // 2) They might try to use the player character for their own purposes --
// > // like aforementioned killing of hated neighbours.
// >
// > That's appropriate behaviour, not intelligent.
// I think it's more intelligent to try to use something at your advantage
// instead of getting rid of it right away `cus it might be dangerous'.
Nope... it's appropriate behaviour, it only seems intellient because
you're looking at it from the monsters view and are thus are blinded to
the big picture. The PC is not a controllable pawn for a monster (it's
less reliable, capable of reversing sides on a whim, and capable of
gaining demigodlike power)... truly intelligent behaviour is to not
trust such a thing, gang up to kill it, and go back to the nice calm
world where all you need to deal with is the usual automatons. What
stops them from doing that and makes them trust or side with the PC...
simply put, it's in character for them to do so (which is why it's
appropriate).
// > // 3) They may underestimate the player character.
// >
// > That's appropriate behaviour, not intelligent.
// This again has nothng to do with intelligence. It's knowledge, and they
// might not have it if there was no way to get it.
The sentient controlling the monster isn't going to underestimate.
The smart play is to be conservative and overestimate. The intelligence
knows the basic rules of the game... it knows what the PC can become,
and it knows the dirt simple approached to stopping that (remove the
things that allow the PC to advance). It's appropriate for the actual
monsters to have imperfect knowledge of things and not understand the
rules of the game (which is a metagame to them). That's not the same
as having them behave intelligently (ie like they're being played
by another person).
// > // 4) They may other goals than just slaying the player character. Living
// > // creatures tend to defend their homes with their lives. I think it's
// > // more likely that they wait, hoping the player character won't come,
// > // and fight when he comes, instead of abandoning all their possesions and
// > // teritory they fought so hard for.
// >
// > That's appropriate behaviour, not intelligent.
// So you call a mania of slaying `a chosen one' intelligent, but a desire to
// enjoy your monster life, do something meaningful, seize control of the
// whole dungeon -- non-intelligent?
Nope... I call those behaviours "appropriate". That's the whole point
of my argument... you're arguing my points, but seem to be completely
blind to it because you're locking the concept of intelligent behaviour
as monsters roleplaying... that's not intelligent, that's approrpiate
behaviour.
// > You seemed to have missed the point (maybe you have it now). Monsters
// > behaving appropriately is not anywhere near the same as intelligently.
//
// Those are two different scales. The monster might behave appropriately
// and intelligently, as well as it might behave appropriately and stupidly.
That's the point. Appropriate monsters are roleplaying... orcs behave
like orcs, if that means doing some stupid tactics in your game world
then that's what they're doing. Fully intelligent monster behaviour
means that they don't behave like orcs, they behave like they're
being played by a sentiant AI.
// > The monsters vastly out number the PC,
// Assuming they unite.
It's the correct move when they're all intelligent.
// > give an actual human control
// > of them to use intelligently against the PC and the PC dies quick and
// > horribly.
//
// And the monsters gain nothing. They are back to their old, boring life.
// Hardly intelligent.
They gain the fact that they stopped the apocalyse! That's hardly
nothing... any monsters who do have plans can now go about them knowing
that there isn't some godlike creature that's going to come stomping
through and destroy everything. That includes Bob the Orc who doesn't
have to worry about the PC juggernaut breaking into his warren and killing
his wife and kids while he's out working in the mines. Sure he's only
making a copper a week at a dull job, but it's not such a bad life in
comparison to living under the sword of PCocles. The desire to live in
"interesting times" (which is an ancient Chinese curse) is foolish and
not intelligent at all, three squares and a peaceful life is far better.
// > I'm all for having monsters behave exactly as your points suggest... and
// > the suggestion that started this subthread can do that nicely without
// > too much work (which is my point... if his point is to do this for
// > some appropriate and quaint monster behaviours, I don't disagree like
// > he implied). Although, there is some diminishing returns that will show
// > up fairly early because a lot of small factors will wash each other out
// > in such a system (more is less/less is more)... there's a balance that
// > needs to be found.
//
// True that this can add flovour to monster behavior. And it's good. It's
// one thing.
// Intelligent behavior is another thing, and it's supposed to make the
// monster (note that it's the monster, not the mosnters) a harder opponent.
// Offcourse he should still behave appropriately at the same time. And
// you don't want all the monsters to be intelligent.
You want hardly any of the monsters to behave intelligently (if any
at all)... appropriate behaviour is the be all and end all here.
The monsters behave the way they should behave, you don't need to bring
intelligence in at all to define it concisely. An "intelligent" monster
that's behaving appropriately is making plans and decisions that are
appropriate to what it is and what it's goals are... those decisions
are not necessarily intelligent, but they should always be appropriate.
// > // 5) It's a cruel world down in the dungeons, and a few hundred dead
// > // monsters along the path of player character is not going to cause a panic
// > // on lower levels -- it's just a norm.
// >
// > But the PC gains power from the blood of the slain. Whereas, he used to
// > susceptible to newts, he's now killing giants (seven with one blow!).
// > Panic! Panic! Everyone for himself! Run to the four corners of the
// > world and tell them to never let one of these things hurt a fly!
//
// Yeah, and those monsters at the gates of hell, will mercifully slay
// the intruder because we asked them to do it. Doubt it.
Where are we going to hell here? This is the monsters splitting up and
running *like* hell (not *to* hell) to warn all the monsters in all the
other games to not do what they did and nip these bizarre outsiders in
the bud.
// There's no cooperation between dungeon lowlife and great archdaemons.
// At least I don't believe in such thing.
Neither do I, because...
// And it's not a question of intelligence.
... that's the point: IT'S A QUESTION OF APPROPRIATENESS!
They don't cooperate because inside the game it's not appropriate for
them to do so. However, it's not intelligent that they don't because
that's the best approach to dealing with a godlike serial murderer.
// How about this scenario:
// I'll gladly give him my life if this will increase his chances of slaying
// that evil overlord that treats all my race as slaves.
//
// It's pretty intelligent too.
The question is whether it's appropriate. If the monster is insane and
far too trusting to martyr itself like this, then it might be, but it's
unlikely that such a situation will make sense inside the game.
Is it intelligent? Not really... you've probably just given the PC
more power to genocide your entire race, even if it is guaranteed that
he's going to go after the overlord eventually (in fact it's more likely
because that means that he'll be hacking his way through them to get to
the overlord (slaves are often the first line because it's easier to
"inspire" them from behind)). There's no intelligence in sacrificing
yourself to the greater of two evils. The best bet is to get rid of
the PC now, and live for another day when we can revolt in a less
risky situation.
In short, you might have an example of appropriate monster metagaming...
the monster is doing something not entirely intelligent but possibly
very appropriate for it /in character/, providing it understands and it
playing the external (to it's POV) roguelike game itself. That
just seems wrong, and I can't get behind that.
Brent Ross
The Sheep
> The Sheep <sh...@sheep.prv.pl> wrote:
> You seem to have completely missed that that's
> what we're talking about. Apporopriate behaviour is different from
> intelligent behaviour because the monsters are acting naturally in their
> own world... sometimes that can be intelligent behaviour from the game's
> point of view as well, but in this case it's not (the monsters are
> far better off combining forces and ganging up on the player... it's
> only that that would be inappropriate for them to do in character that
> prevents that).
Ok, this turns into a really silly discussion. We're repeating ourselves.
I'll try to say what I mean once more again, sorry if it won't
clarify things.
I agree that the monsters should behave appropriately, with and additional
note that sometimes the appropriate behavior for given monster requires
it to behave somewhat intelligently.
The thing I addresed was:
You cite that `grab everything and run downstairs' behavior third or
fourth time on this group, claiming it's the best the monsters can do.
I don't agree with it.
It's surely the worse thing for the player character they might do, but
it's not the best for the monsters.
The monsters might not care about `the apocalypse' and about stopping it,
and it's not a question of their intelligence. It's -- as you said -- a
question of their roleplaying.
For some reason you say that intelligence implies an obsession to kill the
player character, sacrificing any social position, terriority and
possesions.
You also assume that intelligence means divine knowledge and even an
ability to predict future.
And you ignore the fact that not each and every creature in the dungeon
has to be intelligent, and that this fact has a large impact on behavior
of the intelligent ones. Suppose, for example, that the monsters on lower
level are stupid and don't want any truces -- then your intelligent
creatures are in hard position, between the hammer and the anvil.
That's all I wanted to say. Thanks for patience of the whole group. ^^)))
Brent Ross
The Sheep <sh...@sheep.prv.pl> wrote:
// > The Sheep <sh...@sheep.prv.pl> wrote:
// > You seem to have completely missed that that's
// > what we're talking about. Apporopriate behaviour is different from
// > intelligent behaviour because the monsters are acting naturally in their
// > own world... sometimes that can be intelligent behaviour from the game's
// > point of view as well, but in this case it's not (the monsters are
// > far better off combining forces and ganging up on the player... it's
// > only that that would be inappropriate for them to do in character that
// > prevents that).
//
// Ok, this turns into a really silly discussion. We're repeating ourselves.
It's silly because we're agreeing. I created this as an ironic argument
against those would purport that the best AI is the one that plays the
game the most intelligently. I (seemingly) take their side and give a
very workable and reliable stratedgy that monsters can use against the
player to assure that the game provides the level of challege they're
demanding. I further stir things up by actively roleplaying the monsters
position like it's somehow a sensible thing for them to be roleplaying
(using the common Munchkin tactic is to come up with extreme coincidences
and explanations to match their twinks). Then, (ideally) I sit back and
watch as they argue against the absurdity, leading them (hopefully) to
the realization that what they really wanted was appropriate behaviour,
not intellent. That's worked far too good on you.
The ultimate end goal is having them eventually realize that there's a
large difference between intelligent monster behaviour and the behaviour
of intelligent monsters. In fact, it's the same difference as the one
between rollplay and roleplay.
// I'll try to say what I mean once more again, sorry if it won't
// clarify things.
Oh, no need for that... I understood you from the start, you're arguing
against the absurdity of the "ideal monster plan" like I intended.
You have that clue, you just keep for getting duped into believing that
I'm somehow in favour of that.
// The thing I addresed was:
// You cite that `grab everything and run downstairs' behavior third or
// fourth time on this group, claiming it's the best the monsters can do.
//
// I don't agree with it.
//
// It's surely the worse thing for the player character they might do, but
// it's not the best for the monsters.
The worst thing for the player is the best thing for the monsters.
It's really a zero sum situation for the monser AI routine... if it
wants to beat the player, it's a min-max two-player game with some hidden
knowledge which can easily be played conservatively against the ultimate
potential (the player certainly has to do this to win, and in this case,
so does the game).
// The monsters might not care about `the apocalypse' and about stopping it,
// and it's not a question of their intelligence. It's -- as you said -- a
// question of their roleplaying.
//
// For some reason you say that intelligence implies an obsession to kill the
// player character, sacrificing any social position, terriority and
// possesions.
Again you're confusing roleplaying with rollplaying here. The monsters
are trying to win the game... the only way they can do that is to
kill the protagonist. A bunch of knicknacks that aren't real, and are
meaningless to possess once the game ends are a resource that can be
spent like candy or abandoned freely by perfectly intelligent monsters.
Once the game ends everything will cease to exist anyways, so short
term thinking is the most intelligent approach.
// You also assume that intelligence means divine knowledge and even an
// ability to predict future.
Not really. What it means is that the monsters know the rules of the game
itself and that it's being played as a game (much like that episode of
ST:TNG with Moriarity). Having true omnipresence and cheating abilities
isn't really required, the monsters only need to know the rules and the
possibilities (like the player).
// And you ignore the fact that not each and every creature in the dungeon
// has to be intelligent, and that this fact has a large impact on behavior
// of the intelligent ones.
That would be the intent of those who purpose the fully intelligent
AI. Even the lowliest of newts should behave in the most tactically
sound and twinky ways. That's what they were asking for: monsters
who never behave in suboptimal ways.
// Suppose, for example, that the monsters on lower
// level are stupid and don't want any truces -- then your intelligent
// creatures are in hard position, between the hammer and the anvil.
Which is appropriate behaviour for them, not the most intelligent.
Brent Ross
The Sheep
Pretty confusing.
>
> // I'll try to say what I mean once more again, sorry if it won't
> // clarify things.
>
> Oh, no need for that... I understood you from the start, you're arguing
> against the absurdity of the "ideal monster plan" like I intended.
> You have that clue, you just keep for getting duped into believing that
> I'm somehow in favour of that.
Not exactly, I beliweve.
That is, I think you made a few mistakes and oversimplifications that
made the `ideal monster plan' silly, but it doesn't need to be that.
Oh, wait, the `ideal mosnter plan' is silly, but using a good,
`intelligent' AI isn't -- and you seem to like to give that example of
yours when somebody mentions such an AI. It doesn't really prove
anything this way, unless the reader gets caught and shifts his attention.
> // The thing I addresed was:
> // You cite that `grab everything and run downstairs' behavior third or
> // fourth time on this group, claiming it's the best the monsters can do.
> //
> // I don't agree with it.
> //
> // It's surely the worse thing for the player character they might do, but
> // it's not the best for the monsters.
>
> The worst thing for the player is the best thing for the monsters.
> It's really a zero sum situation for the monser AI routine... if it
> wants to beat the player, it's a min-max two-player game with some hidden
> knowledge which can easily be played conservatively against the ultimate
> potential (the player certainly has to do this to win, and in this case,
> so does the game).
I'm surprised you actually play any roguelike games if you view them this
way. I mean the game itself doesn't play agianst you -- it's as if you
said that playing any game according to custom rules is wrong, because
you can always set the rules so that one player will always win. Yes, you
can, but somehow you don't.
If you want to talk about games in the sense of the game theory, you have
to put a thick line between the game rules and the game AI.
I consider the game's world, the dungeon generator, the combat system,
etc, ect. the rules of the game. Also, the monster's limitations are
a part of the game's rules -- the fact that Rogue's orc will first run for
any gold there's in the room, the fact that animals won't use weapons
and armors, the fact that the mosnters of different races cannot
communicate, the fact that Angband's `gray mushroom patch' won't move
around, the fact that the shrieker muchroom with shriek at the sight of
the player, the fact that some monsters will panic in some situations,
AND THE FACT THE MONSTERS HAVE DIFFERENT GOALS. That's part of the rules.
Now, what's left for the `opponent'? The mosnters can choose an action
*from the ones available for them* (made available by aforementioned
rules) so that they will try best to fullfil their goals.
Now, some of those goals might collide with the goal of the player, then
the AI works directly as an opponent (but onl yas long, as the rules
allow). Some of them may be neutral or even help the player -- then,
according to the rules, the monster's AI becomes neutral or even forms
a truce with the player. So it's hardly a two-sided game.
Then again, you may be able to `win' the game without conflicting with
any of the monsters, and then it's not even a game in the sense of the
games theory -- it's just gambling.
> // The monsters might not care about `the apocalypse' and about stopping it,
> // and it's not a question of their intelligence. It's -- as you said -- a
> // question of their roleplaying.
> //
> // For some reason you say that intelligence implies an obsession to kill the
> // player character, sacrificing any social position, terriority and
> // possesions.
>
> Again you're confusing roleplaying with rollplaying here. The monsters
> are trying to win the game... the only way they can do that is to
> kill the protagonist. A bunch of knicknacks that aren't real, and are
> meaningless to possess once the game ends are a resource that can be
> spent like candy or abandoned freely by perfectly intelligent monsters.
> Once the game ends everything will cease to exist anyways, so short
> term thinking is the most intelligent approach.
The rules of the game say that the outcome is computed as if the game
world continued after finishing the game.
> // You also assume that intelligence means divine knowledge and even an
> // ability to predict future.
>
> Not really. What it means is that the monsters know the rules of the game
> itself and that it's being played as a game (much like that episode of
> ST:TNG with Moriarity).
Dunno what you're talking about.
> Having true omnipresence and cheating abilities
> isn't really required, the monsters only need to know the rules and the
> possibilities (like the player).
> // has to be intelligent, and that this fact has a large impact on behavior
> // of the intelligent ones.
>
> That would be the intent of those who purpose the fully intelligent
> AI. Even the lowliest of newts should behave in the most tactically
> sound and twinky ways. That's what they were asking for: monsters
> who never behave in suboptimal ways.
Who are you talking about? Who are those misterious `they'?
Nobody wants to play a game with the rules set so that you always lose.
The game AI, no matter how intelligent must obey the rules.
> // Suppose, for example, that the monsters on lower
> // level are stupid and don't want any truces -- then your intelligent
> // creatures are in hard position, between the hammer and the anvil.
>
> Which is appropriate behaviour for them, not the most intelligent.
What behavior? For whom?
I can hardly define `being in position' as behavior.
Your `example' only say that you can set ridiculous set of rules for your
game and that this game won't be playable then. And it doesn't prove
anything.
There's still lots of room for the `Ultimate AI' in roguelikes.
Brent Ross
// That is, I think you made a few mistakes and oversimplifications that
// made the `ideal monster plan' silly, but it doesn't need to be that.
// Oh, wait, the `ideal mosnter plan' is silly, but using a good,
// `intelligent' AI isn't -- and you seem to like to give that example of
// yours when somebody mentions such an AI. It doesn't really prove
// anything this way, unless the reader gets caught and shifts his attention.
It does. They wanted the monster to work in a perfectly optimal way,
and this is the perfect optimal way for the monsters to accomplish that.
It will, without a doubt, achieve the goal of having the monsters having
the best chance of defeating the player. It's everything they were
looking for.
// > // The thing I addresed was:
// > // You cite that `grab everything and run downstairs' behavior third or
// > // fourth time on this group, claiming it's the best the monsters can do.
// > //
// > // I don't agree with it.
// > //
// > // It's surely the worse thing for the player character they might do, but
// > // it's not the best for the monsters.
// >
// > The worst thing for the player is the best thing for the monsters.
// > It's really a zero sum situation for the monser AI routine... if it
// > wants to beat the player, it's a min-max two-player game with some hidden
// > knowledge which can easily be played conservatively against the ultimate
// > potential (the player certainly has to do this to win, and in this case,
// > so does the game).
//
// I'm surprised you actually play any roguelike games if you view them this
// way.
Boy... you went and got yourself duped again. And this time I wasn't
even trying. Please understand this: I'm talking about having optimal
monster intelligence here... monsters who are playing the game as much
as the player is, and possibly more so because they are playing like
Munchkins and trying to win. That's exactly what this arguement is
against: perfect monster play... it's what was being asked for at
the time. I certainly don't view RLs this way unless someone actually
writes one this way.
// I mean the game itself doesn't play agianst you -- it's as if you
// said that playing any game according to custom rules is wrong, because
// you can always set the rules so that one player will always win. Yes, you
// can, but somehow you don't.
Custom rules? They don't have anything to do with this. The monsters
are playing the same game as the player, and in this case they're united
against him trying to win (because there is no individual win conditions
among them without roleplaying and this is rollplaying). That makes it
one player against a completely united team (the enemy of my enemy is
my friend, and without roleplaying that's absolute in this case).
// If you want to talk about games in the sense of the game theory, you have
// to put a thick line between the game rules and the game AI.
AIs shouldn't have a thick line between them and the rules... they
should know the rules to the game they're playing... they can't really
be independant (that's like the player playing with his eyes shut).
// I consider the game's world, the dungeon generator, the combat system,
// etc, ect. the rules of the game. Also, the monster's limitations are
// a part of the game's rules -- the fact that Rogue's orc will first run for
// any gold there's in the room, the fact that animals won't use weapons
// and armors, the fact that the mosnters of different races cannot
// communicate, the fact that Angband's `gray mushroom patch' won't move
// around, the fact that the shrieker muchroom with shriek at the sight of
// the player, the fact that some monsters will panic in some situations,
// AND THE FACT THE MONSTERS HAVE DIFFERENT GOALS. That's part of the rules.
Having different goals is appropriate behaviour... not intelligent, and
not part of the rules. The Rogue Orc only has to move towards the
gold because it's coded to be an appropriate behvaiour for Orcs... as
far as the optimal intelligent behaviour that was requested, such
a thing is a design mistake.
// Now, what's left for the `opponent'? The mosnters can choose an action
// *from the ones available for them* (made available by aforementioned
// rules) so that they will try best to fullfil their goals.
You're confusing the rules of the game world (the (meta)physics) with the
rules as implemented. The rule that makes the Orc move to gold isn't
a part of the game world... it's merely a coded peice of appropriate
behaviour. I certainly don't assume that Rogue's Orcs are hysterically
drawn towards gold and can't perform any other action... it's simply
appropriate behaviour that they move towards goal and choose not to do
other actions. It's just abstracted in a simple fashion thats functional
for the sense of playability of the game, combined with the appropriate
quaint suboptimal choices of the monster. Not intelligent, but
appropriate.
Mushrooms not moving is a better example because a phyical limitation is
a game rule... but that just means that the monsters need to keep the
dungeon clean of those at all times to maintain their strategy. Thus,
the dungeon is regularly washed every day to prevent mushrooms from
growing anywhere except for the mushroom farms (which are kept well out
of the way, and well down near the bottom of the dungeon). Hey, I
imagine that most forms of mushrooms might well have been made extinct
in a world filled with intelligent monsters fearing the next coming
of the PC.
// Now, some of those goals might collide with the goal of the player, then
// the AI works directly as an opponent (but onl yas long, as the rules
// allow). Some of them may be neutral or even help the player -- then,
// according to the rules, the monster's AI becomes neutral or even forms
// a truce with the player. So it's hardly a two-sided game.
True, if you're not looking for optimal intelligent play. Then silly
roleplaying details like motivations go out the window... the monsters
in the rollplaying, trying to win game don't bother with things like
that anymore than the most munchkin of players. And thus, it's natural
for them to do extreme things like my plan, because Munchkins twink
for extremes.
// Then again, you may be able to `win' the game without conflicting with
// any of the monsters, and then it's not even a game in the sense of the
// games theory -- it's just gambling.
You seem to have this idea that game theory is something about abstract
two-player games. It's not really that at all... it's a good part of
economics and not much about games at all. It really does have an awful
lot to say about such systems providing there's some way to measure
what a "win" is... you can't escape it this way. Besides, death is
almost always a loss for the player, and the strategy of pull back the
weak units to avoid increasing his power while we move the big units
in to gang up on him for the win is a pretty solid way to get that
condition. Other big monsters would naturally be hiding out in large
gangs around any other potential winning goal just in case. In fact,
the monsters probably just reverse the depth concept of their dungeon...
the newts hide protected deep in the bowels of the earth and the
Dragons live in large communes in the entrance rooms. This is how
we protect ourselves from the PC menace.
// > // The monsters might not care about `the apocalypse' and about stopping it,
// > // and it's not a question of their intelligence. It's -- as you said -- a
// > // question of their roleplaying.
// > //
// > // For some reason you say that intelligence implies an obsession to kill the
// > // player character, sacrificing any social position, terriority and
// > // possesions.
// >
// > Again you're confusing roleplaying with rollplaying here. The monsters
// > are trying to win the game... the only way they can do that is to
// > kill the protagonist. A bunch of knicknacks that aren't real, and are
// > meaningless to possess once the game ends are a resource that can be
// > spent like candy or abandoned freely by perfectly intelligent monsters.
// > Once the game ends everything will cease to exist anyways, so short
// > term thinking is the most intelligent approach.
//
// The rules of the game say that the outcome is computed as if the game
// world continued after finishing the game.
That's the "never playing as if it's the last turn" house rule for games
you need to end early. It's never a rule of any game... in fact it's
a good part of some and others have balance measures against it. In
the case of a RL, the win for the monsters doesn't have any special
rules or conditions to stop them from always playing like there's
nothing after the game (after all, the palyer in question ceases to
exist when the game ends as well... what does the game care?).
// > // You also assume that intelligence means divine knowledge and even an
// > // ability to predict future.
// >
// > Not really. What it means is that the monsters know the rules of the game
// > itself and that it's being played as a game (much like that episode of
// > ST:TNG with Moriarity).
//
// Dunno what you're talking about.
Data was using his knowledge of Holmes to instantly solve holodeck
cases. Jordie and Polaski decided to make the game "fair" by having
the computer generate an opponent that could defeat /Data/ (not
just Holmes). The end result is that the computer generated a
sentient character based on Moriarity who knew it was a hologram
and on a starship... because those are the metarules of Data, not
just the rules of Data in the holodeck playing Holmes.
Perfect intelligent play is the same... it's the monsters playing
what to them is a metagame. That's very wrong from a game design
point of view, and the example shows that.
// > Having true omnipresence and cheating abilities
// > isn't really required, the monsters only need to know the rules and the
// > possibilities (like the player).
//
// > // has to be intelligent, and that this fact has a large impact on behavior
// > // of the intelligent ones.
// >
// > That would be the intent of those who purpose the fully intelligent
// > AI. Even the lowliest of newts should behave in the most tactically
// > sound and twinky ways. That's what they were asking for: monsters
// > who never behave in suboptimal ways.
//
// Who are you talking about? Who are those misterious `they'?
The people who were yammering about it years ago when I created the
example in the first place. They were upset with mosnters behaving
stupidly, and were practically demanding that monsters behave as
intellegent as themselves so as to be a challenge. I provided
the monsters behaving intelligently to defeat them... it's inevitable,
the monsters must behave suboptimal, or must be very weak, or must
be few and far between. Intelligent monsters are far too powerful.
// Nobody wants to play a game with the rules set so that you always lose.
// The game AI, no matter how intelligent must obey the rules.
You're still confusing "rules of the world" with what's implemented for
the sake of appropriate behaviour. Just because I code a game where
elven archers never use swords, doesn't mean that elves who are archers
in that world *cannot* phyically use a sword as a *rule*. Inside the
game world, they're actively *choosing* not to wield swords because
that's the *appropriate* behaviour I've choosen for them. An AI that
plays with appropriate conditions limiting it is not playing optimal
and thus not intelligently (it could well be that said elven archer is
out of arrows and standing on a sword of slay-PC... intellegent choice
is to wield it, an appropriate choice is that elven archers never use
swords on principle). It is only playing appropriate to the defined
behaviour of the monsters.
// > // Suppose, for example, that the monsters on lower
// > // level are stupid and don't want any truces -- then your intelligent
// > // creatures are in hard position, between the hammer and the anvil.
// >
// > Which is appropriate behaviour for them, but not the most intelligent.
//
// What behavior? For whom?
The monsters acting stupid because that's their appropriate behaviour
pattern... not wanting truces, because it's appropriate that they don't
want to co-operate for game world reasons (not metagame play reasons).
// Your `example' only say that you can set ridiculous set of rules for your
// game and that this game won't be playable then. And it doesn't prove
// anything.
I'm not setting any rules. The monster have free choice of their
actions and are trying to act optimally. This is what was called for:
monsters equal to players. So what you have is that the monsters are
under the same restrictions as the player. In most games there's no a
lot of restriction put on what the player can do (it's a RL hallmark).
So unless you want Orc PCs to automatically walk to the first pile of
gold coins and sit there forever, it has to be the case that the Orc
doesn't have to do that if it doesn't want to either (and in that case
it only does it if it's not a suboptimal move). What you're describing
as rules goes directly against what was originally asked for... your
appropriate monster behaviours will make the monsters behave suboptimal
and make them easier targets for the player.
// There's still lots of room for the `Ultimate AI' in roguelikes.
I think I've pretty much covered it as far as anyone would want to go.
The point is that that's not what people really want... what you want
is a game that's balanced and playable (thus if they're not abusable,
they better be weak relative to the player or few and far between the
ones that aren't)... while avoiding having the monsters accidentially
do things that are completely pessimal and embarrassing to the designer
(if it's intention design there's no embarrassment).
Brent Ross
Jeff Lait
>
> [1] I've always been against trying for fully intelligent monster
> behaviour... truly intelligent monsters evacuate the level and/or take
> advantage of their massive numbers to overwelm the player, and different AIs
> just result in what we around here call "Monsters behave stupidly in a
> different way" (after the Angband option). I'm an advocate for "monsters
> behave appropriately"... orcs should behave like orcs, it's that simple.
I disagree with your analysis. Many games allow monsters to gain
experience and level up. It would thus seem that the most logical
behaviour is for the monsters of lower level to sacrifice themselves
to higher level monsters to power up those monsters.
In ADOM, where the more monsters of a type die, the more powerful that
monster type becomes, it would be logical to continuously summon and
kill one's own spiders so as to make them infinitely powerful.
It is not just the PC that draws power from blood.
I think in these discussions you should make it clear earlier that you
see the opponent as a *single* sentient being controlling *all* of the
monsters.
--
Jeff Lait
(POWDER: http://www.zincland.com/powder)
Brent Ross
Jeff Lait <torespon...@hotmail.com> wrote:
// bwr...@csclub.uwaterloo.ca (Brent Ross) wrote in message
// news:<cl6e2j$jup$1...@rumours.uwaterloo.ca>...
// >
// > [1] I've always been against trying for fully intelligent monster
// > behaviour... truly intelligent monsters evacuate the level and/or take
// > advantage of their massive numbers to overwelm the player, and different AIs
// > just result in what we around here call "Monsters behave stupidly in a
// > different way" (after the Angband option). I'm an advocate for "monsters
// > behave appropriately"... orcs should behave like orcs, it's that simple.
//
// I disagree with your analysis. Many games allow monsters to gain
// experience and level up. It would thus seem that the most logical
// behaviour is for the monsters of lower level to sacrifice themselves
// to higher level monsters to power up those monsters.
//
// In ADOM, where the more monsters of a type die, the more powerful that
// monster type becomes, it would be logical to continuously summon and
// kill one's own spiders so as to make them infinitely powerful.
//
// It is not just the PC that draws power from blood.
This doesn't invalidate the initial scheme in those games... a game
where the player could gain experience from just surviving might be
a problem for it (the monsters would have to act quickly). You are
very correct in that some games provide additional schemes that the
monsters can use to protect themselves from the dreaded PC.
In short, I don't see where you're disagreement is... you've provided a
second scheme that will work in a game like ADOM. The original scheme
can also work very well in ADOM (and you haven't said otherwise... so
only your first sentence is confusing me), but will also work in
Nethack, Crawl, and Moria/Angand... providing the monsters get the same
capabilities as the player (namely, using stairs).
// I think in these discussions you should make it clear earlier that you
// see the opponent as a *single* sentient being controlling *all* of the
// monsters.
It was a footnote to a far earlier argument... only the result
mattered here.
Still, I don't see the opponent as a single sentient... or at least
I don't have to initially. Given that the monsters are behaving
fully intelligently they know that the other monsters are behaving
likewise... end result is like a classic logic puzzle: they can collude
without any centralized control because they all come to the same
conclusions and know that this is the case. It's simply inevitable
that it ends up the same as the game playing directly against the player
because fully intelligent monsters instincitively have a hivemind effect.
Brent Ross
Chris Reuter
Brent Ross <bwr...@csclub.uwaterloo.ca> wrote:
>In article <0fjcd.6552$NX5....@newsread3.news.atl.earthlink.net>,
>crichmon <crich...@hotmail.com> wrote:
>// What is "surreal time"? The documentation that I found had all text removed
>// that refered to the topic.
>
>Sureal time, as I remember, was a system where the game worked as a single
>player game until two PCs got within each others "sphere of influence",
>at which point the game would move into a real time mode. It was an
>attempt at a compromise.
Hmmm. I have an idea.
The rule is, no two players are ever on the same level at the same
time. And if they are, the game makes copies of the level for each
player and merges them back together when they leave. The in-game
excuse for all of this is that they they were actually there at
different times (in the game) and just managed to miss each other.
I call it: "inconvenient time".
--Chris
--
Chris Reuter http://www.blit.ca
"This guy is nuts. Clearly, you should hit him in the head until the stupid
falls out."
--Elizabeth D. Brooks, <3FCCC23E...@comcast.net>
Brent Ross
Chris Reuter <cgre...@csclub.uwaterloo.ca> wrote:
// In article <cksgto$jba$1...@rumours.uwaterloo.ca>,
// Brent Ross <bwr...@csclub.uwaterloo.ca> wrote:
// >In article <0fjcd.6552$NX5....@newsread3.news.atl.earthlink.net>,
// >crichmon <crich...@hotmail.com> wrote:
//
// >// What is "surreal time"? The documentation that I found had all text removed
// >// that refered to the topic.
// >
// >Sureal time, as I remember, was a system where the game worked as a single
// >player game until two PCs got within each others "sphere of influence",
// >at which point the game would move into a real time mode. It was an
// >attempt at a compromise.
//
// Hmmm. I have an idea.
//
// The rule is, no two players are ever on the same level at the same
// time. And if they are, the game makes copies of the level for each
// player and merges them back together when they leave. The in-game
// excuse for all of this is that they they were actually there at
// different times (in the game) and just managed to miss each other.
Interesting. Doesn't allow for direct interaction, but does allow for
shared experience. Still, you could be players dropping stuff from
an experienced character for another character they just started to
pick up... or perhaps creating a string of characters to rush down to
a problem level with a dangerous monster on it... one will eventually
kill it and the real character can then go through. So it needs some
development work, but is interesting.
// I call it: "inconvenient time".
I'd have been tempted to call is "version control system merge" time.
Brent Ross
R. Dan Henry
The Sheep
I once fancied such a thing, but to be played over e-mail -- you know,
you don't have to be online all the time if there's no interaction.
You only download the changes in levels.
SZDev - Slash
news:cl8l73$132$1...@rumours.uwaterloo.ca...
SNIP
>
> Is it intelligent? Not really... you've probably just given the PC
> more power to genocide your entire race, even if it is guaranteed that
> he's going to go after the overlord eventually (in fact it's more likely
> because that means that he'll be hacking his way through them to get to
> the overlord (slaves are often the first line because it's easier to
> "inspire" them from behind)). There's no intelligence in sacrificing
> yourself to the greater of two evils. The best bet is to get rid of
> the PC now, and live for another day when we can revolt in a less
> risky situation.
>
> In short, you might have an example of appropriate monster metagaming...
> the monster is doing something not entirely intelligent but possibly
> very appropriate for it /in character/, providing it understands and it
> playing the external (to it's POV) roguelike game itself. That
> just seems wrong, and I can't get behind that.
>
> Brent Ross
"... you, and your fellow angels are the last hope for this world, as you
have been given the power to brought individuals into the world, your sons
will be gifted by the Archangels and their lifes will outline over the lifes
of the rest of people in the world ..."
[Guardian Angel manual, v 0.0.001]
Orcs should behave like semi-stupid orcs, Horses should behave using their
instincts, Humans should behave as their position in the social systems
makes them do; a shoopkeeper must sells his wares and supply from someone, a
farmer must go and farm everyday, a wife must stay at home and clean his
house... they really don't take too many decitions, nor they care to change
the fate of their world, their AI must be plain Appropiate.
However, there are special individuals out there, which goal is not to keep
up with their lifes but to become powerful, smarter and gain enough support
to save the world, these are the sons of the Angels, the player controls
one, but there are others in the world too, looking to comply the same goal
of the player, they may even meet and cooperate. Some of them are controlled
by an AI module, which must get near to Intelligent, one of them is
controlled by the most advanced intelligence module, which interacts with
the game through the UserInterface, he is the candidate with more
possibilities to sucess in the task, tough he must contact the other sons of
the Angels, as otherwise his task is doomed.
--
SZDev - Slash
Slashing, the Outcast Dragon of the -={UDIC}=-
Weblog: http://www.livejournal.com/users/szdev
Website: http://szdev.cjb.net
Konijn
Care to explain this huge problem in a sense that it is a problem for the players ?
Cheers,
T.
>
> Brent Ross
Brent Ross
Konijn <kon...@gmail.com> wrote:
// bwr...@csclub.uwaterloo.ca (Brent Ross) wrote in message
// news:<cl6ukn$vge$1...@rumours.uwaterloo.ca>...
// > In article <cl6pip$2dv$1...@news.vol.cz>, konijn_
// <kende...@hotmail.com> wrote:
// > //
// such a fashion
// > // to play an online rogue game, I would say that is a success, not a
// failure.
// >
// > Actually, in this case it's a huge failure. The players are cooperating
// > not to *play* the game, but to intentionally *break* the game. Having
// > the players distorting the nature of space time is not a success,
// > it's a huge problem.
//
// Care to explain this huge problem in a sense that it is a problem for
// the players ?
It makes the game boring because it's trivially exploitable. Games with
bad holes like this tend to have a short play life before they get
shelved.
Brent Ross
SZDev - Slash
news:fvs9lc...@catarneh.blit.ca...
> In article <cksgto$jba$1...@rumours.uwaterloo.ca>,
> Brent Ross <bwr...@csclub.uwaterloo.ca> wrote:
> >In article <0fjcd.6552$NX5....@newsread3.news.atl.earthlink.net>,
> >crichmon <crich...@hotmail.com> wrote:
>
> >// What is "surreal time"? The documentation that I found had all text
removed
> >// that refered to the topic.
> >
> >Sureal time, as I remember, was a system where the game worked as a
single
> >player game until two PCs got within each others "sphere of influence",
> >at which point the game would move into a real time mode. It was an
> >attempt at a compromise.
>
> Hmmm. I have an idea.
>
> The rule is, no two players are ever on the same level at the same
> time. And if they are, the game makes copies of the level for each
> player and merges them back together when they leave. The in-game
> excuse for all of this is that they they were actually there at
> different times (in the game) and just managed to miss each other.
>
It is a good idea as it solves most of the problems, but it washes the
experience of "Let's form a gang to kill this powerful Balrog at 4500 ft",
which is one of the most attracting features, after all, why play a
multiplayer game if you will be alone most of the time???
I guess some of the lonely RL players would like the indirect interaction
tough...
> I call it: "inconvenient time".
:D
>
>
> --Chris
>
>
> --
> Chris Reuter
http://www.blit.ca
> "This guy is nuts. Clearly, you should hit him in the head until the
stupid
> falls out."
> --Elizabeth D. Brooks, <3FCCC23E...@comcast.net>
--
Konijn
> My objection is to the idea that group abuse equals success.
> Realism doesn't have much to do with that (the game is plain
> broken)
Broken in my dictionary ( and the dictionary of WoTC ) is when a
feature/option/choice is _always_ chosen by the player no matter the
drawbacks.
Prime example : buying a whip in older angbands, no way you could beat
2 or 3 attacks with 1D6 damage, so every char had one if the player
knows about this.
Prime example : their latest banned card in Mirrodin, every deck
played with it, no matter color or genre.
Now,
with somewhat improved AI the advantages of this system are to me so
small, that not all players ( for sure not me, probably not you )
would go for it.
Ergo-> not for sure broken, it would depend on the implementation of
the AI.
... although the game also has realism stretched beyond
> any reasonable bound (some abstraction is reasonable, but some
> sense of realism[3] is always a requirement).
I dont buy it, not to mention that this fatigues me.
Timed Chess = abstracted battle of 2 armies with some absurd rules (
king dead->you loose), realism totally destroyed by 1 army managing to
move units in seconds while the other seems to take minutes. And this
is still fun to play. I dont care about realism, I care about fun in
all my games.
Anyway, let's stay friendly neighbours and agree that we disagree on
brokennes, fun, stretched realism beyond any reasonable bound and
requirements of sense of realism.
Cheers,
T.
Isaac Kuo
>Hmmm. I have an idea.
>The rule is, no two players are ever on the same level at the same
>time. And if they are, the game makes copies of the level for each
>player and merges them back together when they leave. The in-game
>excuse for all of this is that they they were actually there at
>different times (in the game) and just managed to miss each other.
>I call it: "inconvenient time".
I had an similar idea based on the way Moria type
roguelikes never send you to the same level twice. The
handwaving justification is that the stairways are
actually confusing mazes or something like that.
Based on this idea, each level has a pool of maps. When
you enter a level, you get sent randomly to one of
the unoccupied maps. If all of the maps are occupied,
a new map gets spawned into the pool.
For continuity purposes, you might make the assignment of
newly entered maps less random. For example, you might
keep track of the following data for every staircase:
1. The destination staircase
2. The last time this staircase was a destination
When a player enters a staircase, the computer checks to
see if the destination staircase's map is occupied.
If not, the player enters the destination map. If so,
the destination staircase is changed. The computer
checks all of the staircases in all of the level's
unoccupied maps (spawning a new map if necessary).
The least recently used staircase is chosen as the new
destination.
Either way, the timestamp of the destination staircase
is updated, and the destination of the destination is
replaced by the starting staircase.
The desired behavior of this algorithm is to encourage
continuity when a player backtracks through a staircase.
For example, consider a situation where there is one
level 1 map, and two level 2 maps (2A and 2B). Two
players, Amy and Bill are playing. The following
sequence of events can occur:
1. Amy enters level 1.
2. Amy goes down the staircase to level 2A. Level 2A's
up staircase gets pointed back to level 1.
3. Bill enters level 1.
4. Bill goes down the staircase, which gets reassigned
to point to 2B. Level 2B's up staircase gets pointed
back to level 1.
5. Amy retreats back up to level 1. Level 1's down
staircase gets repointed back to level 2A.
6. Amy goes back down to level 2A.
7. Bill retreats back up to level 1. Level 1's down
staircase gets repointed back to level 2B.
8. Bill goes back down to level 2B.
From the point of view of either player, everything looks
consistent. There is a minimum amount of inconsistency
generated, based on the fact that both Amy and Bill inhabited
level 2 at the same time. They never inhabited level 1 at
the same time, so there was never a need to spawn a new
level 1 map.
Now here's the cool part--what if level 1 had had two
different down staircases? In this case, it's possible
for there to be no inconsistency at all! Suppose Amy
took down staircase "Alpha", while Bill took down staircase
"Beta". In that case, down staircase Alpha points to
level 2A while down staircase Beta points to level 2B.
With this algorithm, it's possible for a network of maps
to interconnect and remain more or less consistent except
as necessary.
I guess you could call it a "Lazy Lonely Hogwarts" scheme.
The staircases shift around to keep everyone lonely--but
they're lazy about it and only shift when necessary.
Of course, just how different is a lonely multiplayer
roguelike gaming experience from a single player roguelike?
Is it really so interesting to enter an already plundered
map? I guess the most interesting aspect is the potential
to plunder player character corpses.
I think the inclusion of communal "town levels" is an
important enhancement. It would allow for trading of
items without the kludginess of staircase-dancing.
Isaac Kuo
Brent Ross
//
// > My objection is to the idea that group abuse equals success.
// > Realism doesn't have much to do with that (the game is plain
// > broken)
//
// Broken in my dictionary ( and the dictionary of WoTC ) is when a
// feature/option/choice is _always_ chosen by the player no matter the
// drawbacks.
I don't go with _always_, since there are always people who will
completely try to avoid the abuse (for example, I've never used an
artefact from a different class in Nethack, and when playing Seiken
Densetsu III I actively avoiding bringing up rings when casting spells to
avoid the chaining abuse). Broken is simply allowing for aberrant player
behaviour to me... when something is making the game unfair, it's
broken. I shouldn't have to avoid the Munchkin abuses to appreciate
the game, it should work well for everyone (nor should I have to contend
with the twinky abuse driven scores and bones files that the Munchkins
will be pumping out).
// Prime example : buying a whip in older angbands, no way you could beat
// 2 or 3 attacks with 1D6 damage, so every char had one if the player
// knows about this.
This is the classic example of what I call aberrant player behaviour.
Naturally, you'd expect the fighter types to by big weapons and the
mages to buy the small... but as you point out, the exact opposite
is the "correct" course of action. You can take you fighter and
buy the biggest weapon you can afford if you want to roleplay that...
but even the native fighters in that world must know they can do far
more damage with a whip. The game's simply wired up backwards.
// Prime example : their latest banned card in Mirrodin, every deck
// played with it, no matter color or genre.
Magic is played a lot by Munchkins. Personally I prefer the idea
of just grabbing my old cards, making a few fixed decks that are
balanced and play well against each other, and just play with
those. The deck building aspect is far too prone to twinkiness.
// Now,
// with somewhat improved AI the advantages of this system are to me so
// small, that not all players ( for sure not me, probably not you )
// would go for it.
Personally, I'd prefer that the system just avoid the questionable unfair
situation to begin with. Limit the number of actions on MGM and keep
different time frames as far apart as possible... no one should ever be
able to affect something in a different time frame than their own. Note
that this has implications beyond maximum spell range and LoS... detection
is also an issue, since an "active" player could use detection spells in
a cleared area and then use an external communication mechanism (talking,
internet messaging, etc) to give information to a "passive" player who's
freezing the time frame in another area (within detection radius).
In short, it doesn't really matter if non-Munchkin players aren't going
to abuse it... nobody should even have to think that this might even be
possible in the game.
// Ergo-> not for sure broken, it would depend on the implementation of
// the AI.
I only see this as trying to sweep the problem under the carpet.
It's trying to add noise and disincentive to cover up a problem which is
probably better avoided in the first place. Focusing on making a game
that is comfortably playable by different players (who react phyically
at different rates) and characters (who have different complexities of
interface) in the same time frame is a better approach (with possible
sureal time when the players are completely separate). I'd rather
see that seriously tried and fully experimented on first before
trying to hack in overly sureal time.
// ... although the game also has realism stretched beyond
// > any reasonable bound (some abstraction is reasonable, but some
// > sense of realism is always a requirement).
//
// I dont buy it, not to mention that this fatigues me.
// Timed Chess = abstracted battle of 2 armies with some absurd rules (
// king dead->you loose), realism totally destroyed by 1 army managing to
// move units in seconds while the other seems to take minutes. And this
// is still fun to play. I dont care about realism, I care about fun in
// all my games.
Chess is an abstract game. Roguelikes are very much not abstract games
which makes realism more of a concern. An abstract RL would certainly
be interesting (although most people would probably say it's not a RL
at all, because themes are so strongly ingrained along with the idea of
roleplaying).
It'd be something like this: I move my piece(s) around the board,
picking up (and perhaps dropping) tokens. I can do things like trade
in a blue token to remove adjacent pieces from the board... or perhaps
move pieces or tokens around in some fashion. Certain squares and
piece/token arrangements on the board may give additional options
or cause immediate effects. I win by achieving a number of points,
getting my piece to a certain square, or collecting certain sets of
tokens or pieces. The computer plays the opponent piece(s) who have
their own winning condition (which may be the same... it depends on
if the game is symetric). There could still be some dice or hidden
information (backgammon and card games are abstract and have these)...
but there wouldn't be any ties with the real world. You can freely
define the rules about blue tokens without having to worry about player
associations involving common sense and reality.
Chess does have it's consistancies (it's not a pure abstract, it does
carry a theme)... wall time and world time aren't related (just like
roguelikes), and the concept of losing the king to losing the game is one
of the strongest realistic features of the game (traditionally, removing
the head of state[1] meant chaos and the ability as a conquering power
to immediately seize control and restore order... that's still the case
in some governments today).
// Anyway, let's stay friendly neighbours and agree that we disagree on
// brokennes, fun, stretched realism beyond any reasonable bound and
// requirements of sense of realism.
I'm not so sure we do differ much on fun. You seem to suggest that
you're willing to avoid abuses for the sake of fun, and so am I (I just
still consider the game broken because I have to do that and can't play
at full potential without worrying).
As for sense of realism, probably not so much there either. I only
require the realism that's appropriate to the game genre and it's design.
For an abstract like Chess, by definition, my expectations are low... I
expect to have to learn all the moves and rules by rote. Something like
a RL is a different matter: they tend to be very complex as far as games
go (consider how long the rules would be... describing Nethack as board
game rules would make the Magic Realm[2] book look like a light read).
In a roguelike, having a good number of things behave in common sense
ways helps keep the game light and reduce fatigue... the key is to
avoid going to far into making the player micromanaging the reality[3].
I don't consider micromanagement to be essential for realism, what I do
require for that is that there's an underlying phyics (and metaphysics)
that the player can grow to understand and being to intuit things about
the game world. This allows for inductive reasoning to play a role (ie
that was an odd, let's run a few experiments and see if we can find out
what happened) as well as reducing the amount of rote learning that's
required (ie the player can make generalized "proverbs" that describe
whole sets of actual rules). When you present a system that appears to
have some common sense behind it, but implement it with results that are
counterintuitive you go against that and add to what must be learned by
rote and cannot be intuited. It complicates the game and enough of that
can lead to fatigue (I don't want to think about then possibility of time
distortions... even if I'm trying to avoid abusing them, there's still
the possibility that my actions inadvertently allow the game to abuse
(and possibly kill) my character because an unfair system can run both
ways).
Brent Ross
[1] Traditionally these are also divine representatives. When the
chosen of God gets removed or killed that's not a good omen.
[2] Magic Realm is an old Avalon Hill roleplaying board game. It came
out at a time when rulebooks were written like lawbooks (ie they were
very completist and explicit with clauses and subclauses). Most people
learn to play Magic Realm from people who learnt from other people.
Only a handful have successful learnt from the rules themselves. You can
find a copy of the rules at boardgamegeek.com under Magic Realm... it's
quite fun to randomly go to one of the eighty odd pages and read a random
passage for a good chuckle. Magic Realm is a lot simpler than Nethack
(than most modern roguelikes actually)... it has to be because board
games can't run the same level of complexity without taking forever.
[3] For example, it's common sense that people need a light source in a cave,
eat food, and sleep regularly... but if those things aren't going to be
real factors in the game, get rid of them. Make the abstraction that the
character has a light source and that food and sleep are all acccounted
for and don't make the player deal with it. For example, in Crawl we
choose not to make light an issue, but do use food as a balance measure.
Thus, eating becomes a controllable part of the game for the player,
but having a lightsource is a common thing that every character manages
on their own. Pick and choose these things carefully so you don't end
up with too many of them and the game doesn't become burdened by
micromanagement.
Isaac Kuo
>"Chris Reuter" <cgre...@csclub.uwaterloo.ca> escribió en el mensaje
>news:fvs9lc...@catarneh.blit.ca...
>>The rule is, no two players are ever on the same level at the same
>>time.
>It is a good idea as it solves most of the problems, but it washes the
>experience of "Let's form a gang to kill this powerful Balrog at 4500 ft",
>which is one of the most attracting features, after all, why play a
>multiplayer game if you will be alone most of the time???
>I guess some of the lonely RL players would like the indirect interaction
>tough...
I think a "pure" lonely multiplayer roguelike eliminates too many
of the rewarding multiplayer experiences. However, I can see a
lot of potential for a nearly pure lonely multiplayer game where
players can trade items.
Now that I think about it, the "let's form a gang" aspect can be
accomplished with a clique based multiplayer system. In this
case, each player has a list of "buddies". A player can enter
an already occupied map if he's buddies with everyone else in
the map. Within each map, the sequence of play is strictly
turn based, perhaps with a timeout failsafe. Perhaps the timeout
value is determined semi-democratically--each player sets his
desired timeout value and the median is chosen.
You can even have different sequencing systems as players desire.
Some players may prefer a strictly turn based system. Other
players may prefer a strictly realtime system.
Isaac Kuo
R. Alan Monroe
>"SZDev - Slash" <ragn...@hotmail.com> wrote in message
>Some players may prefer a strictly turn based system. Other
>players may prefer a strictly realtime system.
This thread got me thinking about how to handle multiple players in a
turn based system. How about this: If you are the first to encounter a
monster, you become the "owner" of that monster. If other players come
into LOS, their attacks have no effect on monsters you own, so there's
no need to fiddle with trying to match turn speeds between the two
players. If you wanted, you could let the newly arrived player swing
at the monster, just for atmosphere, but he would have a 100% miss
rate.
Alan
Ray Dillinger
> This thread got me thinking about how to handle multiple players in a
> turn based system. How about this: If you are the first to encounter a
> monster, you become the "owner" of that monster. If other players come
> into LOS, their attacks have no effect on monsters you own, so there's
> no need to fiddle with trying to match turn speeds between the two
> players. If you wanted, you could let the newly arrived player swing
> at the monster, just for atmosphere, but he would have a 100% miss
> rate.
It would be not very fun to get beaten to death by a monster
you can't touch.
And if the monster can't touch you either, I envision hordes
of munchkins having their fastest guy take point so he sees
all the monsters first, and then runs away, while the invulnerable
munchkins walk in and vacuum up all the stuff.
Bear
Gerry Quinn
bwr...@csclub.uwaterloo.ca says...
> // Why do you always assume that:
> // 1) All the monsters should cooperate?
>
> Because the monsters are "perfectly intelligent" like the players (and
> some designers) ask for. As such, they realize that the most dangerous
> thing to their regular life is the PC... this monster will exponentially
> grow in power if it's allowed to kill any of us!
Monsters that breed explosively in Angband should help the player. He
will kill lots of them, but even more will grow if they are near him.
If the green worm masses play their cards right, by the time the player
goes downstairs half the level will be filled with them!
- Gerry Quinn
Jeff Lait
> In article <774acfb8.04102...@posting.google.com>,
> Jeff Lait <torespon...@hotmail.com> wrote:
> // I think in these discussions you should make it clear earlier that you
> // see the opponent as a *single* sentient being controlling *all* of the
> // monsters.
>
> It was a footnote to a far earlier argument... only the result
> mattered here.
I think, however, it is the core of the argument!
> Still, I don't see the opponent as a single sentient... or at least
> I don't have to initially. Given that the monsters are behaving
> fully intelligently they know that the other monsters are behaving
> likewise... end result is like a classic logic puzzle: they can collude
> without any centralized control because they all come to the same
> conclusions and know that this is the case. It's simply inevitable
> that it ends up the same as the game playing directly against the player
> because fully intelligent monsters instincitively have a hivemind effect.
They can only collude without central control if there is perfect
information. Each monster sees a different world, so may come to
different conclusions about what optimal behaviour would be. "We seem
to be working at cross purposes here" I do not think it is all
apparent that there is One True Strategy for dealing with the PC, so
it is possible that the monsters would choose different incompatible
strategies.
This is why I think your example works best if you phrase it in terms
of Computer Vs Human rather than Monsters Vs PC. The latter argument
seamlessly ties into the concept of what you call appropriate
behaviour. Namely, "How would a sentient player play from the
viewpoint of a monster". This is not necessarily uber-intelligent
munchkin behaviour. Plenty of presumeably sentient people play in
suboptimal manners.
Brent Ross
// news:<cl9vjg$s2u$1...@rumours.uwaterloo.ca>...
// > In article <774acfb8.04102...@posting.google.com>,
// > Jeff Lait <torespon...@hotmail.com> wrote:
// > Still, I don't see the opponent as a single sentient... or at least
// > I don't have to initially. Given that the monsters are behaving
// > fully intelligently they know that the other monsters are behaving
// > likewise... end result is like a classic logic puzzle: they can collude
// > without any centralized control because they all come to the same
// > conclusions and know that this is the case. It's simply inevitable
// > that it ends up the same as the game playing directly against the player
// > because fully intelligent monsters instincitively have a hivemind effect.
//
// They can only collude without central control if there is perfect
// information. Each monster sees a different world, so may come to
// different conclusions about what optimal behaviour would be. "We seem
// to be working at cross purposes here" I do not think it is all
// apparent that there is One True Strategy for dealing with the PC,
Not really... "I if I am weak, dive down as fast as possible to get behind
big monsters.... they'll be racing upwards to kill the PC before he can
develop into a threat". Monsters just need to evaluate their chances
against a low level PC and react immediately when the shouting starts.
They don't need to have anything more complicated for a general strategy.
Once the big guys get at the player, then you might have a point as they
may have different tactics that require resolution (berserkers keeping the
fire lines open for the mage artillery), but that's tactics not strategy.
// This is why I think your example works best if you phrase it in terms
// of Computer Vs Human rather than Monsters Vs PC. The latter argument
// seamlessly ties into the concept of what you call appropriate
// behaviour.
My point here is more that becomes like a single computer player
is playing directly against the player if the monsters behave perfectly
intelligently. The standard logic puzzle assumpts basically give the
monsters the ability to perfectly see what the other monsters will
do. They could do slightly better if the Computer is allowed to
fully co-ordinate the pieces, but they don't have to... the little
monsters just need to run away warning everyone, and the big monsters
just need to rush in for the easy kill. The monsters can decide on
that individually without having to be controlled by a hivemind.
Which brings us back to this point:
// so it is possible that the monsters would choose different incompatible
// strategies.
They certainly wouldn't do that... they know exactly how the other
monsters are going to react to the PC being in the dungeon and thus
the last thing they would choose is a strategy that's incompatible and
would foobar the defense of the dungeon... and knowing that the other
monsters are intelligent, they know that none of them will choose an
incompatible strategy either. So they'll all end up choosing something
simple (like up/down) knowing that it's effective, dirt simple to the
point of foolproof, and the other monsters will all come to the exact
same evaluation (because they're all logical thinkers... they don't
really see the world differently at all). If different strategies were
chosen, they'd either be completely compatible (thus forming a greater
metastrategy), or low-risk, easily abortable (thus not conflicting).
Monsters who did otherwise would not be intelligent (and thus don't
exist in this dungeon).
// Namely, "How would a sentient player play from the
// viewpoint of a monster". This is not necessarily uber-intelligent
// munchkin behaviour. Plenty of presumeably sentient people play in
// suboptimal manners.
Yes, but they only have control of a single being in the game world.
The odds are badly stacked against the PC (give a human player control
of a group of orcs and they'll crush a PC unless the PC is far too
advanced... at which point they could probably get away with minimal
casualities if they didn't choose to gamble on ogging)... what helps to
balance things is having the monsters pretend that they actually have real
lives (ie roleplaying their part), rather than simply eliminating the
PC threat at all costs (ie munchkin powergaming). Having the monsters
played in a munchkin manner, using the twinkiness of numbers and home
field advantage just leads to quick PC death.
Brent Ross
Brendan Guild
[snip]
> Not really... "I if I am weak, dive down as fast as possible to get
> behind big monsters.... they'll be racing upwards to kill the PC
> before he can develop into a threat". Monsters just need to evaluate
> their chances against a low level PC and react immediately when the
> shouting starts. They don't need to have anything more complicated for
> a general strategy. Once the big guys get at the player, then you
> might have a point as they may have different tactics that require
> resolution (berserkers keeping the fire lines open for the mage
> artillery), but that's tactics not strategy.
In that case it seems to me that when people ask for sentient monsters,
what they want is sentience in tactics, not strategy. They'd probably be
happy for the monster to behave perfectly intelligently in a fight and
otherwise just sit around like an inanimate object to be discovered by
the PC. That seems like a viewpoint that solves everything nicely,
doesn't it?
Brent Ross
Brendan Guild <do...@spam.me> wrote:
// Brent Ross wrote:
//
// [snip]
//
// > Not really... "I if I am weak, dive down as fast as possible to get
// > behind big monsters.... they'll be racing upwards to kill the PC
// > before he can develop into a threat". Monsters just need to evaluate
// > their chances against a low level PC and react immediately when the
// > shouting starts. They don't need to have anything more complicated for
// > a general strategy. Once the big guys get at the player, then you
// > might have a point as they may have different tactics that require
// > resolution (berserkers keeping the fire lines open for the mage
// > artillery), but that's tactics not strategy.
//
// In that case it seems to me that when people ask for sentient monsters,
// what they want is sentience in tactics, not strategy. They'd probably be
// happy for the monster to behave perfectly intelligently in a fight and
// otherwise just sit around like an inanimate object to be discovered by
// the PC. That seems like a viewpoint that solves everything nicely,
// doesn't it?
Not really. Weak monsters would realize that they were outclassed
and run away screaming, driving all the other monsters off the level.
Monsters with detection and groups with scouts could easily use these
tactics... groups of orcs might sacrifice one member and the PC would
never see any other orcs. Eventually the PC will be bored or die of
starvation.
As I said before, give me control of a group of orcs vs a PC and
the situation is either a surely dead PC or minimal orc casualities,
depending solely on the level of the character. In order for the PC to
have a chance against a group of orcs, they need to behave incompetently
so the player has an advantage he can exploit. Fortunately, this is
in character for orcs... but it's not in character for all monsters.
The AI doesn't even have to be that impressive to lopside the game...
I've messed around with the monster AI in Crawl, and the end result
is that centaurs become sure death traps[1] for PCs with nothing more
than hold position and fire tactics. With perfectly intelligent monster
tactics you either need very few monsters using them, design monsters to
be relatively weak in comparison to the PC, or have the player control
a party of PCs (which is pulling away from a RL and towards a CRPG).
Brent Ross
[1] Not quite 100%... but it essentially makes most classes and styles
completely unplayable, which is undesirable. The balance is really
quite tricky... even orc wizards and priests are extremely lethal in
very short order when they start with straight, uninterupted barrages.
Trust me, it's a good thing from a player enjoyment and playability
standpoint if the monsters are simply forbidden from using the same cheesy
tactics as the player (just think about how boring it must be from the
monsters point of view to watch Magic Missile after Magic Missile being
cast every round like clockwork).
Chris Reuter
The Sheep <sh...@sheep.prv.pl> wrote:
>> // I call it: "inconvenient time".
>>
>> I'd have been tempted to call is "version control system merge" time.
>
>I once fancied such a thing, but to be played over e-mail -- you know,
>you don't have to be online all the time if there's no interaction.
>You only download the changes in levels.
Whereas I thought of doing it over the web. The client is written in
Flash, JavaScript or maybe a Java applet (although those seem a wee
bit clunky) and talks to the server by XMLRPC or some kind of
HTTP-based thing.
The nice thing is that players can jump right in without needing to
download anything. On the minus side, ewww--Flash!
--Chris
--
Chris Reuter http://www.blit.ca
"The trouble with the world is that the stupid are cocksure and the intelligent
full of doubt."
--Bertrand Russell
Mike Blackney
>
> In that case it seems to me that when people ask for sentient
> monsters, what they want is sentience in tactics, not strategy.
> They'd probably be happy for the monster to behave perfectly
> intelligently in a fight and otherwise just sit around like an
> inanimate object to be discovered by the PC. That seems like a
> viewpoint that solves everything nicely, doesn't it?
Why would you want perfect intelligence? When I play multiplayer games
I don't even get that! If I throw a steak on the ground, I expect that
the charging wolves will go for it instead of me.
But failing that, I think that when people ask for sentient monsters
they'll usually be happy with monsters that don't act like algorithms.
--
michaelblackney at hotmail dot com
http://aburatan.sourceforge.net/
Latest version 0.95 2-5-4
Martin Read
>But failing that, I think that when people ask for sentient monsters
>they'll usually be happy with monsters that don't act like algorithms.
Well, unless you have access to hardware random numbers, they *have* to
act like algorithms. Very complicated algorithms, perhaps, but
algorithms nevertheless.
--
Martin Read - my opinions are my own. share them if you wish.
Mike Blackney
> "Mike Blackney" <michael...@hotmail.com> wrote:
> > But failing that, I think that when people ask for sentient
> > monsters they'll usually be happy with monsters that don't act
> > like algorithms.
>
> Well, unless you have access to hardware random numbers, they
> *have* to act like algorithms. Very complicated algorithms,
> perhaps, but algorithms nevertheless.
In the same way that a mother who travels through time assasinating
American presidents is acting like a mother. A very complicated
mother, perhaps.
Chris Reuter
SZDev - Slash <ragn...@hotmail.com> wrote:
>"Chris Reuter" <cgre...@csclub.uwaterloo.ca> escribió en el mensaje
>> The rule is, no two players are ever on the same level at the same
>> time. And if they are, the game makes copies of the level for each
>> player and merges them back together when they leave. The in-game
>> excuse for all of this is that they they were actually there at
>> different times (in the game) and just managed to miss each other.
>It is a good idea as it solves most of the problems, but it washes the
>experience of "Let's form a gang to kill this powerful Balrog at 4500 ft",
>which is one of the most attracting features, after all, why play a
>multiplayer game if you will be alone most of the time???
You could let the players form clans. Give each clan a clubhouse and
communal bank account and players be able to store their stuff there
as well as contribute items to a communal stash. Have lots of quests
that require several different missions, each requiring different
character classes.
The clubhouse could be realtime. So could one or more town levels
where there's no combat allowed. That lets people interact in
realtime. You could also build IM capabilities into the client,
although it might be just as easy to use one of the existing services.
Dead players could haunt the clubhouse as ghosts until somebody
resurrects them.
Of course, if you really care about ganging up on a monster, you're
pretty much stuck going realtime.
--Chris
--
Chris Reuter http://www.blit.ca
"To me, one of the most disappointing things about babies is the way they
refuse to meekly accept anything their parent decides is good for them. I
assume this changes later on." --Jeff Vogel
Isaac Kuo
>In article <2tsoa1F...@uni-berlin.de>,
>SZDev - Slash <ragn...@hotmail.com> wrote:
>>"Chris Reuter" <cgre...@csclub.uwaterloo.ca> escribió en el mensaje
>>>The rule is, no two players are ever on the same level at
>>>the same time. [...]
>>It is a good idea as it solves most of the problems, but
>>it washes the experience of "Let's form a gang to kill
>>this powerful Balrog at 4500 ft", which is one of the
>>most attracting features, after all, why play a
>>multiplayer game if you will be alone most of the time???
>You could let the players form clans. Give each clan a
>clubhouse and communal bank account and players be able
>to store their stuff there as well as contribute items
>to a communal stash. Have lots of quests that require
>several different missions, each requiring different
>character classes.
>The clubhouse could be realtime. So could one or more town levels
>where there's no combat allowed. That lets people interact in
>realtime.
I think "free-for-all" would work better than realtime.
A free-for-all system lets any player move at any time,
limited only by how quickly or slowly he types.
I consider "realtime" to mean something where each turn
takes a particular amount of time, on average.
>You could also build IM capabilities into the client,
>although it might be just as easy to use one of the
>existing services.
You should build chat capabilities into the client, so
the players don't need to switch windows constantly to
go between IM'ing and controling his character.
>Dead players could haunt the clubhouse as ghosts until somebody
>resurrects them.
>Of course, if you really care about ganging up on a monster, you're
>pretty much stuck going realtime.
I disagree--I think plain old turn based is the best, if
you limit things to the trusted clan. In this case, the
rule is:
No two players from different clans are ever on the
same level at the same time.
Keeping everything strictly turn based is simple, elegant,
and intuitive. There are only two disadvantages to a strict
turn based system:
1. A single player can halt the game leading to excessive
waiting.
and
2. When there are too many players, the waiting time is
excessive even when each player isn't being purposefully
slow.
Both of these problems are mitigated if strict turn based
play is limited to trusted players.
With a realtime system, you have problems also. There are
actually several ways to accomplish "realtime" which are
different in important ways.
For example, a "strict" realtime system means a particular
player's turn occurs strictly at the scheduled time. A
commanded action doesn't occur immediately upon the final
keypress, but rather when the next scheduled time occurs.
This can have bad gameplay effects if the player's scheduled
time occurs immediately after the scheduled time of a nearby
monster--the player needs to "guess" what move he'll need
to make beforehand. The obvious solution to this problem
is to ensure that the monsters always move immediately after
the players. However, this still leaves the problem of
players stumbling with each other. The player who is "first"
in line gets to see his situation before choosing his action,
but the player who is "last" in line may end up attacking
thin air or bumping into another player or unable to react
to a teleported monster.
A more flexible realtime system lets a player's move occur
immediately upon his last keypress, but he has to wait a
particular amount of time before making another move. This
system is more intuitive and feels more responsive than a
strict realtime system. However, there's a question of when
the monsters move. Depending on how monster timing is done,
the results may be confusing.
A less confusing realtime system would be based on time
windows. Every n seconds, the monsters move and a new time
window starts. Each player is allowed to make one move
during a time window, which occurs immediately upon his last
keypress. If a player completes no command within a time
window, a "rest" command is assumed.
A hybrid turn based system can use this time window concept.
The monsters move, starting a new time window. However, the
duration of this time window is flexible. If all players
make a move, the time window immediately ends and the monsters
make their next move. If not, a preset maximum duration
provides timeout failsafe in case a player stalls.
The realtime time window system has the advantage that it
sets up a regular tempo for the players to acclimate to.
The turn based time window system has the advantage that
the game tempo can speed up or slow down as desired.
Surreal time takes the time window system to the next level
by overlapping the time windows. This leads to weird effects
as players can get ahead/behind and sometimes you have to
wait for other players to "catch up".
Isaac Kuo