Can you tell if you're playing a computer?
Tom Ewall
class level, including chess, Scrabble and backgammon. In chess it's
fairly easy to tell if you're playing a computer, because they play
moves a human would never think of. Similarly in Scrabble, you can
tell. In both Scrabble and chess, even if you set the level so it's
not so strong, you can still tell, because it will play some great
moves a human wouldn't think of and make some moves so stupid a human
wouldn't make them.
My question is if you (who's ever reading this) can tell if you're
playing a computer when playing backgammon.
Warwick
I don't think so Tom. Some people believe that a consistent lag in making
moves is an indication. The bots dont support inline analysis of play so
one assumes your opp has to manually enter the game state into the bot he
is using. This means delays in moves that seem completely straighforward,
and should require little thought or time to play.
The best thing to do is rely on post match analysis, if the opponent
consistently comes up as ET then that opponent is either collaborating or a
genuine world class player. Both are good reasons not the play them again
(for money any rate).
best
Warwick
J
Just personal experience - not much but my tuppence worth ...
If the whole match is played by computer it's easy to tell i think. Humans
have plans - bot's don't apart from simple blitzing,priming and a few other
things.
It's when someone pauses for a bit too long on a cube they should really
know or something like that you start to wonder. Guess that's called 'bot
assisted' play.
My only wonder is ... with so many people playing exactly like bots in the
hope that it will win more ... maybe it is hard to tell after all.
But to answer your question ...
If you gave me a bot to play at random online and it played the whole match,
yeah i could tell everytime.
"Tom Ewall" <tew...@lycos.com> wrote in message
news:85bf428.03122...@posting.google.com...
Tom Ewall
with on-line cheating.
I guess one thing that comes to mind is that computers don't make
errors. When I go over games, my most common mistakes (at least of
the blunder type) are not doing things I know just by not noticing
(like missing a blot that could be hit, forgetting the cube is on my
side, not running a guy to safety -- that sort of thing), and
computers would never make that sort of mistake. But apart from that,
I don't notice the type of thing you're mentioning -- not playing with
a play. But it could be I'm not sure what to look for.
Could you say some more about what makes it evident to you that you're
playing a computer? I'd be interested.
I'll speak for myself, regarding playing without a plan. I'll have a
certain plan, but at each moment, I'm asking myself, what's the best
way of playing a certain roll. If the position changes, then I try to
change my plan, if I think that's appropriate. One of things I try to
constantly ask myself is if the plan I was employing is still
appropriate, especially when something important happens (a back man
escapes, a man gets hit, saftied, etc.). I would think that this sort
of playing, which any reasonable player should do, would be similar to
the playing without a plan play of the computer you describe.
"J" <blah...@hotmail.com> wrote in message news:<3fed0...@mk-nntp-2.news.uk.tiscali.com>...
CLohf
>My question is if you (who's ever reading this) can tell if you're
>playing a computer when playing backgammon.
Computers tend to play hopeless positions differently from humans. For example,
if it is impossible to save a gammon the computer will evaluate all possible
moves as equal and pick one. Humans, in contrast, continue to make 'good' moves
in hopeless situations.
Albert Silver
I don't know about Scrabble, but you're quite right about chess.
Still, even then it's sometimes a tough call. As to backgammon, I
really don't see it. Of course, you can tell provided you are a
considerably stronger player than your opponent and your opponent is
quite weak (in theory), but even so, I don't find it so obvious,
unless it's over more than one match. Playing a strong expert and a
bot is a very similar experience, and I don't see how you're going to
identify a few imprecisions over the course of a match compared to the
bots' even fewer imprecisions.
I will say this much comparing playing programs in chess and
backgammon: it is far more enjoyable to play them in backgammon in my
opinion. Chess has that wonderful unique item known as the
combination, and the programs can certainly shine at times there, but
you also know that you are a colossal favorite to lose, and badly,
because it is at a super grandmaster level, with no dice to save you,
and even post-mortem analysis may not help you properly raise your
game. What I mean by this, is that it will tell you the best move (or
one of the best) and give you a 15-ply deep variation to justify it.
While interesting, you also know there's no way you can reproduce that
15-ply line in your games, so even though you might indeed find that
same best move, you'll probably need the kind of instruction only a
master or grandmaster can provide with reasons and plans. Mind you, I
love chess, and love to play it, but when talking about playing
against programs, that is my view. I have a number of chess programs
that I use to study and analyze with, but I don't really play them
anymore.
Albert Silver
Dick Adams
> ...
> My question is if you (who's ever reading this) can tell if you're
> playing a computer when playing backgammon.
On FIBS, computers can't kibitz coherently. However, I have suspected
players of using bots - generally because they take so much time before
rolling and before moving. To verify this, save the match, analyze it
using GnuBG, then review the doubling situations. If you are being
doubled consistently on the first D/P opportunity, your opponent could
be using a bot.
On the other hand, I was thrashed by a player who consistently doubled
early and GnuBG reported my opponent as "awful" at cube play but "Take
your dice to Vegas" for lucky rolls.
Dick
Tom Ewall
backgammon with the computer a lot more fun than playing chess. I was
trying to figure out why, and two thoughts came to mind. One is that
I'm not so aware I'm playing a computer as far as the play itself is
concerned. Secondly you have a decent chance of winning.
Regarding the why of a move, I think the same thing applies as in
chess. That is, as to why a given move is better than another, to
really understand it, you would need grandmaster really figure out why
as the methods the computer uses to evaluate moves are too complex to
be helpful over the board. On the other hand, I think playing the
computer can definately help your intuition as to what moves are good
in certain situations and I find playing the computer has helped my
game a lot. I can't say the same amout chess.
In Scrabble, the computer will do things like play a vowell next to a
triple letter score, so that by playing "XI" you can score 48 points.
People don't do that. On the other hand, it might play a word like
"YEUK" which lower ranking scrabble people wouldn't do either.
silver...@hotmail.com (Albert Silver) wrote in message news:<f9846eb9.03123...@posting.google.com>...
Douglas Zare
Dick Adams wrote:
> tew...@lycos.com (Tom Ewall) wrote:
>
> > ...
> > My question is if you (who's ever reading this) can tell if you're
> > playing a computer when playing backgammon.
It is usually very easy for a strong player to tell that not all of
his/her opponent's moves are made by a computer. It just takes
one mistake (or one good play) that a bot would never make.
Against a very strong player, it can be tough to tell in a short
match or session. Post-mortem analysis with a bot usually
makes it clear that the opponent played with a very different
style from the top bots, which tend to agree with each other
more than they agree with the best human players.
> On FIBS, computers can't kibitz coherently. However, I have suspected
> players of using bots - generally because they take so much time before
> rolling and before moving. To verify this, save the match, analyze it
> using GnuBG, then review the doubling situations. If you are being
> doubled consistently on the first D/P opportunity, your opponent could
> be using a bot.
Did you mean D/T? I disagree that this is a good thing to look
at to try to distinguish human from bot play. Doubling
aggressively is easily learned from a bot. It is not uncommon
for competitive players to play a long match with no missed
doubles.
The types of errors made by current bots (whether told to add
noise or not) are very different from the types of errors made
by humans. Bots rarely overlook plays such as switching points
in a blitz. Humans often do. Bots will rarely make huge
blunders. Even the top players in the world still make
occasional blunders larger than 0.400 EMG, and many more
checker play mistakes worth more than 0.200 than bots do,
for example, in mutually advanced positions.
Douglas Zare
J
Here are some examples anyway.
You find yourself in a position with 1 man back on the 24. the bot is
towering itslef up as you build your board. it will keep doing this until it
rolls a joker. The bot defence by waiting for the lucky roll is not always
best some say. All i know is when playing a human, even if he decides that
bot style moves the whole way here are correct, his pauses in angst at
having to do them will tell me he's human.
Switch roles in this scenario. You are the guy with the towers and the bot
has the lone man on the 24. The bot will never advance this man with a 2
unless it absolutely has to since usually it means a drop in equity. It
doesn't see the future gain in equity in the race yet, even on 3ply
sometimes although a human will see it immediately - even in a situation
like this the way the bot seems to be applying a fudge in its evaluation
will tell me if i'm playing a bot or a human.
How about another favourite of mine, the backgame and its defence. The bot
defence initially relies on some strange notion that more checkers caputured
leads to more gammons and backgammons. Completely wrong. Humans just don't
make these mistakes. Correction - they do :-) Gnu seems to be getting a bit
better at backgames and its defence but it's still appalling at both - to me
anyway. There's the classic (if you can call it that) problem where bots
have not one clue about containment in some situations with many checkers
back. Sure these positions are rare - but the cube handling by humans (even
error prone ones like me) will be completely and i mean completely different
from that of a bot in these games - it won't be perfect by a long chalk -
but it will be better.
Ok another situation then ...
these ones are a little more vague. I don't know a whole load about bots but
one thing i notice is situations where it takes a little longer for its
'race engine' to take over from its 'contact' engine than is really
neccessary. A human is capable of completely changing their strategy for the
better - sometimes it takes the bot just one roll longer. I'd post a few up
but can't be bothered who reads any of this anyway. They all involve
completely giving up one idea and changing to another in one roll. bots
can't do this so well.
And there are more ... these all involve bearin. My bearoff is a little
error prone but i'm happy with my bear *in*. (most the time anyway). Bots
have some strange ideas about bearing in but seem to be pretty good now.
It's harder to tell here but used to be easy. there is still a slight
difference in style though.
Hmm anything else? oh yeah, bots are crap at money games. i mean crap. :-)
J.
"Murat Kalinyaprak" <mu...@compuplus.net> wrote in message
news:2831c30c.04010...@posting.google.com...
> Douglas Zare wrote 3FF33096...@math.columbia.edu
>
> >>> My question is if you (who's ever reading this) can tell
> >>> if you're playing a computer when playing backgammon.
>
> > It is usually very easy for a strong player to tell that
> > not all of his/her opponent's moves are made by a computer.
>
> Perhaps only if his opponent is not another strong player..!?
>
> Obviously a clown like Zare couldn't, but do you think that
> for example a Woolsy or a Magriel could tell the difference
> between Jelly Grendell and Jerry Fish...? Har har... :))
>
> > It just takes one mistake (or one good play) that a bot
> > would never make.
>
> Anybody who is at least half-intelligent can easily see
> that Zare inserts "or one good play" in parantheses just
> to balance bull-shit with horse-shit...! :)
>
> Of course, he is more than welcome to shove this statement
> of mine up my ass by giving a few examples of better moves
> that humans could but bots couldn't make... Of course, in
> attempting to do this, I would like to see what he would
> use as a measuring stick...?? :))
>
> > The types of errors made by current bots (whether told to
> > add noise or not) are very different from the types of
> > errors made by humans. Bots rarely overlook plays such as
> > switching points in a blitz. Humans often do. Bots will
> > rarely make huge blunders. Even the top players in the
> > world still make occasional blunders larger than 0.400 EMG,
> > and many more checker play mistakes worth more than 0.200
> > than bots do, for example, in mutually advanced positions.
>
> Bla bla bla bull-shit... :( I have just posted 25 matches
> of 7 points here between myself (i.e. "nobody" within the
> incestuous gamblers-anonymous circle of sick scum) and the
> gnudung. From the authoritatively confident voice in your
> spewing phd-level bullshit here, it sould be very easy for
> you to underline in some of those matches what you are
> balbbering about...!?
>
> I'll bet my ass that you can't and that's why I think you
> are nothing more than an inconsistent/senile mathemagician
> clown... :((
>
> MK
Paul Tanenbaum
> class level, including chess, Scrabble and backgammon. In chess it's
> fairly easy to tell if you're playing a computer, because they play
> moves a human would never think of. Similarly in Scrabble, you can
> My question is if you can tell if you're
> playing a computer when playing backgammon.
Not backgammon, but related:
http://daily.stanford.edu/tempo?page=content&id=12643&repository=0001_article
An interesting and active research problem.
---
Paul T.