do you think a computer can play poker?

[blaze666]

although most people would say that a computer cannot possible play poker, and all the books say that as well, i would have thought that poker would have been the perfect game for a computer to win at. this is because [online] poker is a game of primarily logic. a computer would be able to calculate pot odds instantly, and it would be able to record every single players hand histories, and be ble to recall tham, and say for example, they will only call x amount, with x hand, with x in the pot, and x probability of winning the hand, against a player who has played x amount of hands. don't you think?

[Zetack]

If a computer can play chess at the highest levels in the world, undoubtedly a computer could play poker at at least an expert level. How hard it would be to program and what kind of computing resources you would need I have no idea.

The important difference it would seem to me, would be that chess is a game of perfect information and poker is a game of imperfect information. Still, you have to think it could be done.

--Zetack.

[blaze666]

well zetack, according to ian pierson, of the 'futureology' department, at BT, the PS3 will be about 1% as powerul as a human brain. by 2050, he predicts that a computer will be able to carry out as many functions as a human brain. although, they thought we'd be going on holiday to the moon by now.

[adamstewart]

[ QUOTE ]
chess is a game of complete information and poker is a game of incomplete information

[/ QUOTE ]


This is a key difference.


(Nevertheless, as per my thread regarding "perfect poker", an extensive database may help to fill in some of the incomplete information, or at least give the most probable possibility).


Adam

[csuf_gambler]

[ QUOTE ]
If a computer can play chess at the highest levels in the world, undoubtedly a computer could play poker at at least an expert level.

[/ QUOTE ]

wrong, do you see why?

[David Steele]

[ QUOTE ]
Quote:
chess is a game of complete information and poker is a game of incomplete information

This is a key difference.

[/ QUOTE ]


It is incomplete for humans too, big deal.

OK you can't use chess algorithms to solve poker
but who thought that anyway?

Also some computers already play pretty well, UofA at
heads up NL etc.

D.

[FishAndChips]

Ah, but computers play chess quite differently than humans do. Tactically computers are almost perfect. They don't blunder away pieces, miss mates in 7, or fall for many stalemate traps etc. However, even the best programs, like Fritz and Shredder, tend to play a strategically inferior game.

By keeping positions "closed", and making moves with long term strategic advantages, Grandmasters can often thwart the computers computational powers with superior strategic play. I honestly don't even believe that what computer programs do is "think" while playing chess. They simply created millions, or billions ,of board positions and score each one. The move that leads to the highest score is made. The programmers have previously done all the programs thinking, and all the program does is essentially calculate. When your TI 85 calculator adds 2+2=4 [img]/images/graemlins/cool.gif[/img] is it "thinking"? But I digress...

Chess programs prove that computers can calculate and play chess very well with a good algorithm, but it also shows how strategically awful they can be in situations that call for the computer to make deductions with imperfect information (that is the computer can't see all moves required for mate and thus makes moves while the result of the game is still uncertain.) This is why getting a program out of it's opening book has often proved a good way to combat it.)

All of these programs have what is known as the horizon effect, and that is the point beyond which it calculates. If a computer makes a move that looks good to its evaluation function that sees 40 moves in advance, but loses after move 41, it doesn't "know better." This horizon effect is essentially the point at which chess becomes a game of imperfect information for the computer, and it is at this point that computers fail miserably.

I think poker computers can be made to play solid poker, but I think they will not be anywhere close to top human players for many years to come. Things like knowing when a player may be on tilt, or when someone is likely to change gears would be tough for a program to adapt to.

There is some credence to what people refer to as "feel" and "instinct" in poker, and when combined with a good knowledge of game theory and mathematics, I believe that a human can play at a higher level than what a computer program can muster. You could write code that tries to cover such things, but humans have thousands of years of evolution that make the psychology of the game second nature, while a computer must be taught all of this within the constraints of a program developed by a select group of programmers etc. I just don't think computers will be there for some time.

Poker offers frequent opportunities (even more so than chess) for computers to have to make deductions in situations that they are unfamiliar with. When this happens their strategic weaknesses become a liability and much of its calculation and statistical power may be nullified. In short, man still rules tha machine. [img]/images/graemlins/cool.gif[/img]

[pokergrader]

Easily. Computers can play at a very high level too.

Go heads up against the UofA poker pot and see how you do: http://www.cs.ualberta.ca/~games/poker/

[pokergrader]

[ QUOTE ]
Snip

[/ QUOTE ]

You make a lot of valid points, but you are missing one obvious point. Making a computer play chess well uses tons of processing and computing power, which given the time constraints leaves very little "extra" time for doing other advanced strategy calculations (even if such a thing existed).

Poker on the other hand is very simple to process. The computer can practically spend all its time categorizing opponents as processing the statistics on the cards is trivial. It can create a moving graph of the aggression of each opponent, and track changes over the long-term or short term. It can detect "bad beats" just as easily and you can and perhaps turn on a tilt checker and check that opponent for drastic changes in play.

I'm not sure exactly how well its been done, but it is clearly possible with todays very powerful computers, you just need a good programmer teamed up with a good poker player.

[Izenra]

Yes, but easy to counter : Track you stats with pokertracker, play a while, and then when you are against the computer, do exactly the opposite of what your poker trackers stats says. The computer is goin to be focked up and will blow.