The suggested format is a transcript for a potential game, annotated with
footnotes. For example:
You find yourself in another  small clearing,
surrounded by stout looking oak trees.
>break off a branch 
 The parser is aware of the number of similar locations
that have been visited, and inserts "another" etc. in the
description as appropriate (see Dynamic Text Generation
for details). The player can also use words like "other"
to refer to things, eg. "go back to the other clearing".
 There is no explicit "branch" object here, but the
parser's world model tells it that branches are parts of
trees (and can be broken off). The world model also just
says there are "trees" here, but after a branch is broken
off, a new tree object (with a broken branch) is created
and placed in this location.
An alternative format to a transcript would be a small essay describing your
innovations for IF; or if you are really feeling brave, a short IF game
demonstrating your ideas. A combination of these formats is OK (eg. an essay
along with a transcript).
Points will be awarded for feasibility, usefulness and originality.
Submissions should be sent to da...@hsa.com.au by April 30th, 2007 (with
"Innovation Comp" in the subject line).
Intents to enter are appreciated but not required. There will be a two week
judging period, during which anyone may vote.
The rejected name makes it sound like you want things *impossible* to
do with current IF systems, but your examples sound more like you're
looking for extensions. (Both of those would be very possible in
Inform 7 at least; in fact I've personally implemented something very
like the first one.)
The "feasibility" requirement also makes it sound like you're looking
for things that *could* be done in current IF systems, but just
haven't been yet.
Sorry for the lack of clarity ...
The focus is on something which is possible to do, given enough time (the
feasibility requirement) - but which does not currently exist, because of
the difficulty of doing it *fully* (or maybe just because nobody thought of
doing it yet. It doesn't really have to be difficult).
The two examples from the original post can both be done less-than-fully at
the moment, but I'm thinking of something more general and complete.
The first example (inserting the word "another" when you come to a second
small clearing) can be done easily enough on a case-by-case basis, but I'm
thinking of an IF system that does such things automatically without the
author writing any special code. The IF system would either need to
understand English (and parse the room description "You find yourself in a
small clearing", identifying the nouns etc. and identifying similarities to
other room descriptions - ugh!), or would require room descriptions to be
entered in a format that it is easier for it to parse and understand. It
would also need to keep the "output" text in sync with the "input" text -
the player needs to be able to refer to "the other clearing", etc. as well.
The second example - allowing the player to refer to a "branch" because
there is a "tree" there, and allowing the PC to break it off - requires a
huge and detailed world model to be done fully and completely. Something
like the "cyc ontology" (see google), a hierarchy of objects a player might
happen to mention in a game (along with their properties, relationships to
each other, appropriate verbs, etc.). There are a lot of potential nouns in
The "innovation" doesn't have to be as difficult as these two ... it may
just be a different approach that nobody has thought of yet. What do you
wish was easier to do (or you got "for free") when you write IF?
I hope that explains the Comp a bit better ... let me know if it still
sounds a bit contradictory,
> The rejected name makes it sound like you want things *impossible* to
> do with current IF systems, but your examples sound more like you're
> looking for extensions.
I suspect that _everything_ can be done with current IF systems, since
both Inform 7 and TADS3 are presumably Turing complete. :)
Very cool idea for a comp, by the way.
Even if they are Turing complete, there are still some things that are
not currently possible and others that are very hard.
For example, suppose you wanted to play IF in Mandarin Chinese.
Dealing with the grammar would be hard; displaying Mandarin fonts is
currently not supported (I think).
Or suppose you wanted to launch a first-person animated sword-fighting
section for all the fight scenes, while having the intervening scenes
as traditional IF.
Or what about a game that NEVER gave error messages, but found the
best fit command for whatever you typed in, allowing for extra words
and spelling errors. That would be fun to design but would take a
different approach to that of the existing systems. It might not
always do what you intended, of course.
You find yourself in another small clearing,
surrounded by stout looking oak trees.
You heave at the tree. A heavy branch hits you on the head and falls
to the ground.
>take the damn branch [ignores extra words]
Taken. No need to swear.
Where the fallen branch dislodged some leaves, you see something
>shiny thing[if only noun, tries the default action, usually
It looks like a key.
>tkae key [copes with spelling errors]
You take the key.
There is a twitch. A rope noose tightens around your ankles and you
are flung violently upwards, head down.
>wjdhfsohjf [nonsense is interpreted as one of the default actions:
looking around, taking inventory, commenting on your progress in the
game, or attempting some reasonable action. In this case, looking
around is chosen]
A forest clearing is above your head and blue sky is below your
feet. Or is that the other way round? You start to feel a little
>cast magic spell on rope [the player is clutching at straws. The
game doesn't recognise 'magic' or 'spell' and thinks 'cast' means
throw. 'cast on' is not supported. The game gathers that the player
is trying to do SOMETHING to the rope, and answers accordingly]
The rope round your ankles is extremely painful, but you can't even
reach it from here. Glancing downwards (upwards?) your heart sinks
(rises?) even further as you notice the rope is fraying. There's a
ten-foot drop (head first) to the ground if it breaks.
>scream aaaaaaa! [just recognising the 'scream' bit]
You hear rustling in the bushes - your screams have attracted
someone. A moment later a large tiger pushes its way into the
clearing and licks its lips.
You are carrying:
- a branch
- a key
- a stun-gun
>I stun the tiger with the stun gun and then fly home. [recognises 'stun tiger with stun gun' and 'fly']
You shoot the stun gun at the tiger. Unfortunately, your aim is poor
(with being upside down) and you miss, dropping the stun gun.
Then you try to fly. It's not as easy as birds make it look.
>stun the tiger again, dammit. [recognises 'stun tiger' and as the stun gun is out of reach, uses the best weapon left]
As the stun gun is lying on the ground, well out of reach, you try
using the branch.
You fling the branch at the tiger's head with more desperation than
accuracy. It misses by yards, but by some fluke strikes the trigger of
the stun gun on the ground, which fires. The dart hits the tiger on
The tiger looks cross, then confused, then sleepy. It slumps over and
starts to emit tiger snores.
Alas, no more tigers appear to have heard you.
The rope emits nasty little fraying sounds.
>I escape [unrecognised, but tries to suggest some useful action]
You try to do something, anything, but it is no use. You drop.
Luckily, your fall is broken by a sleeping tiger.
The brief excerpt reminded me of Matt Fendahleen's "Time
Bastard" (Walkthrough Comp).
> ... what about a game that NEVER gave error messages, but found the
> best fit command for whatever you typed in, allowing for extra words
> and spelling errors. That would be fun to design but would take a
> different approach to that of the existing systems. It might not
> always do what you intended, of course.
[... transcript ...]
Your Innovation Comp entry is a little early :-)
But I enjoyed it!
So I win if no-one else enters? That's two votes for me already ;-)
Are multiple entries allowed, by the way?
Yes, multiple entries are fine ... if you are overflowing with ideas, that
might be a better idea than putting them all in a single transcript.
I'll weigh in here... This is not real amenable to demonstration
in a spurious transcript, but here's the notion:
The IF whirligig tracks the attitudes, game-knowledge and -beliefs,
and likely emotional state of the player.
Let's break that down some:
Beliefs about the game: What the player understands to be the
case in the game world, and what the player understands the
rules to be.
Knowledge about the game: The whirligig (wg) compares the
player's beliefs with the actual state and rules of the game.
Player attitudes: The wg takes notes on how the player seems
to respond to situations, and makes guesses as to the attitudes
that these responses reflect. Does the player attack NPC's for
no particular reason? etc.
Player emotional state: What NPCs and outcomes has the
player formed an attachment to? What are the player's priorities
I have my doubts that this is possible, at least with the current
state of the art; probably requires some AIish programming that
is frankly outside my bailiwick. However:
An IFwg that studied and built up an understanding of the player's
mind could then find ways to motivate the player real-time: crude
The player is motivated by revenge? Have the villian hurt him or
cheat him, and then give the player the opportunity to retaliate.
The player is motivated by romance or friendship? Have the villian
abduct the heroine or friend.
The player is motivated by acquisition? The villian owns vast
quantities of ill-gotten gold.
And so on.
These are better motivators than the DOOMish "you need to find the
blue key" jazz, because those are external motivators, while these
are internal motivators that have more meaning to the players.
In this way, the IFwg can now play with the player on a responsive
level, where it is *actively* tracking the player's movement through
his own inner emotional space; and it can adapt the world to suit
(This is the advantage of rpgs over writing, btw; a good DM can
make a kick-ass adventure with a little insight into his player's
minds, whereas a writer does not have personal access to his
audience, has to do more work, and has to make educated guesses
about their responses.)
People have written about making IFwgs that can adapt the world
pro-actively, to cope with player innovation, rather than just
with the rigidly coded structures: I'd argue that this, having the
psych the player, is a necessary precondition to having a wg that
can adapt to the player intelligently.
Interesting thoughts - I agree that some kind of model of the player is
necessary if you want the game to adapt to their playing style, etc. This
would also be a very different kind of game to traditional IF ...
A challenge for you: Can you come up with a transcript (or several parallel
transcripts) that involves a concrete situation that would be affected by
this approach? A few reasons for exploring a transcript are:
(1) It makes the idea clearer to others - "Here is how this might work in a
(2) It demonstrates the difference it would actually make to a IF game - the
idea comes closer to reality, instead of remaining just a theory.
(3) It highlights practical implementation issues - given a concrete game
situation, an IF author can think about what it might take to implement the
idea (or at least part of it).
As you say, the problem is a bit "AIish" ... a reason for asking for
transcripts rather than actual games in the Innovation Comp is so that
people can present ideas that might otherwise take years of work to
demonstrate. A concrete transcript makes the "feasibility" side a bit
clearer - what would it take to code this particular example (in a general
way, not just as a special case)?
> A challenge for you: Can you come up with a transcript (or several parallel
> transcripts) that involves a concrete situation that would be affected by
> this approach?
I dunno that I want to take you up on that challenge... Lemme mull it
You already gave a general example of how it could work in a game with
your ideas of how to alter the circumstances surrounding a villain. I
think it wouldn't be that hard to write up a possible transcript
involving those ideas.
Also, would any of the psych tracking be visible to the player? Would
there be any indication of what the game *thinks* the player is
feeling, other than the changes in the game structure? That is, in
just a transcript, it seems that it would be a little difficult to
show what your "whirligig" is doing as opposed to normal changes and
effects on the structure of the game as the result of the player's
actions. You'd have to have multiple transcripts at the very least,
and probably comments inserted as to why certain things happened as