Slashdot Mirror
Researchers "Solve" Texas Hold'Em, Create Perfect Robotic Player
Jason Koebler writes The best limit Texas Hold'Em poker player in the world is a robot. Given enough hands, it will never, ever lose, regardless of what its opponent does or which cards it is dealt. Researchers at the University of Alberta essentially "brute forced" the game of limit poker, in which there are roughly 3 x 10^14 possible decisions. Cepheus runs through a massive table of all of these possible permutations of the game—the table itself is 11 terabytes of data—and decides what the best move is, regardless of opponent.
...they got banned by 6241 online casinos and bragging here is the only thing left?
Wouldn't another robot which knows of all possible decisions of this particular robot be better that this "Perfect Robotic Player"?
In theory there is no difference between theory and practice. In practice there is. - Yogi Berra
The computer cannot win every hand, which means it must lose some hands. Since it cannot control how large the bet gets, and in real gambling there is no such thing as infinite reserves, then the computer is still subject to the same worries the pros have: whether you can weather the losses and not go bankrupt long enough for your skill to have you come out on top eventually.
Recently I noticed that Texas Hold'em is only half of the game. The betting is the real strategic part. Unless the bot can do this well, I don't it will ever really "beat" a human player.
There is another computer that triples the bet every time it loses and leaves when it wins and says something about knowing when to walk away.
And then there is the computer that deals the cards and takes 4% of the pot every round, muttering "House always wins".
The third computer that can beat it keeps giving it complimentary alcoholic beverages until it gets a buffer overrun.
The fourth computer doesn't "believe in no win situations" and reprograms the computer to lose, but then Spock finds out and he gets kicked out of the Academy.
Priest: "Universe from nothing, no laws of physics, sped up time"+ huge discrepancies. Creationism? No. Big Bang Theory
I'm a former professional poker player, now semi-pro and working again in the IT industry. In a game like poker, to "solve" the game, from a mathemartcal and game theory point of view, means to develop a strategy that is "unexploitable", which basically means "mistake free". If two game-theory perfect players were to play against each other, then their "expectation" would be zero, as if they were flipping a coin between each other. Neither would make a mistake, so only te randomness of the cards would determine the winner of a given hand. In the long run, both perfect players would win as often as they lose.
But in a real poker game, human players make lots of mistakes. A player who adjusts their strategy to exploit these mistakes will win vastly more than this (formerly theoretical) "perfect player". The game-theory optimal strategy is focused on not losing, rather than exploiting mistakes and winning the most.
So in an actual game, the expert human player will outperform the computer because the other humans in the game are exploitable.
In live play, especially in tournaments, computer solutions are used in poker. In particular, when the game is "heads up" (only two players), and the chips are not deep, which happens at the end of every tournament, then the correct strategy is to "jam or fold" all hands. The solution to this has been determined in a computer and top players have the table memorized.
If this subject interests you, I HIGHLY recommend "The Mathematics of Poker", by Chen and Ankenman.
...does it know when to fold 'em? When to walk away? When to run?
Koans and fables for the software engineer
You've got to lookup when to hold 'em
Lookup when to fold 'em
Lookup when to walk away
Lookup when to run
You never lookup your money
When you're sittin' at the lookup table
There'll be time enough for lookups
When the dealin's done
You can better understand what is going on by considering the much simpler game Rock paper scissors. 'Perfect' here basically means the strategy gives you the best possible worst case.
For RPS, the perfect strategy (using the term in the same sense as it is used for the poker bot) is to play completely randomly. There is no way to gain an edge over this strategy, no counter-strategy which will give you more than 50% chance of winning, even if you know your opponent's strategy. (In this case, there is also no strategy which will give you less than 50% chance of winning against the 'perfect' strategy.)
For the poker bot, there is no strategy that will give you greater than 50% chance of winning against it in a two player game. If you know its strategy perfectly (but of course you don't know its cards) the best you can do is to equal that 50% chance (which is what happens if it plays itself.) Unlike RPS, you can can lose to the perfect poker bot by playing poorly. Also, as noted in the article, the perfect poker bot always plays as if it were playing against perfect opposition. A good human player will fleece you faster then the perfect bot, because the human player will notice your peculiar imperfections and exploit them, choosing to play in a way which would be suboptimal against a perfect opponent, but superior against you.
Quattuor res in hoc mundo sanctae sunt: libri, liberi, libertas et liberalitas.
Comment removed based on user account deletion
even in a fair game, the house wins.
Not always. Poker isn't played against the house; it's played against other players. The house just gets about 4 percent of any pots that flop.
Blackjack is the other casino game that can be "beaten". It's played against the house, but the house plays like a robot, and its only advantage is that it takes double busts (hands where both the the player's cards and the dealer's cards total more than 21). The player has plenty of advantages, including double payout for player A-10, the "insurance" side bet on dealer A-10, split A and 8, double down on 10 or 11, and standing on less than 17. There's a basic strategy that by itself makes the house advantage negligible, and if you can mentally estimate when A-10 is more likely, you can know when to bet higher and when to insure. This was enough for the MIT Blackjack Team to turn a modest profit.
Comment removed based on user account deletion
Poker, for humans, is luck and psychology in the main. It's about convincing the other guy that your hand is better/worse than it is.
But the problem is that if you're playing against a robot, he doesn't care what you want him to think. He knows exactly what the odds are of you having any particular card, and what that means in terms of him beating you. Playing the odds will win on average. It's how it works. It's how casinos get rich enough to have marble floors and air conditioning in the middle of the desert.
If you don't get this, you'll lose a lot of money playing against this machine.
But no doubt you have a "system". Or you can "read" players.
Expert human poker players have a good enough knowledge to know the odds (even if only approximate) on every turn of the card, but they can't analyse every possible combination in time. The rest of it is trying to "lie" to another human. It's rare for a poker player to be the best poker player consistently and for years, precisely because its not as simple as having skill, but overwhelmingly a good amount of luck.
Otherwise, sorry, but anyone on planet would be rich by just plugging in what cards they were given on PartyPoker into an app that tells them the percentage chance of winning. On those kinds of site (last time I used it) there was no human interaction enough to perform any kind of psychology, so it's entirely skill of the game and luck of the cards. And if you can eliminate the need for skill of the game, then by your theory you'd win (almost) every time. You don't. And poker-playing bots only make money when playing against imperfect humans. Play them against each other and you'll be there forever as the money goes back and forth, back and forth (subject to game rules such as blinds, etc.).
Poker has the "most" skill of any casino game. In any of its variants. But that's not a lot. Claiming that a skilled player would beat a bot hands-down? Strange that the poker sites are so hot on blocking bots, then, isn't it?
Bluffing in poker is only relevant in order to make a human opponent make an irrational decision. That's what you're trying to achieve. If the human makes the rational decision every time, then it comes down to luck alone. Making a rational decision every time involves a hell of a lot of card-counting and knowledge of the odds, so few can actually do it properly (I'm a mathematician, I wouldn't dare state that I could calculate the odds without perfect knowledge and a lot of time).
But no amount of bluffing changes the cards in your hand, the cards left in the deck, what the next card will be, or what your opponent probably has in their hand.
The reason people enjoy poker is because a good player can trick a bad player into playing worse. A computer program like this isn't subject to such tricks.
Prove me wrong. Play a statistically-significant number of games against the thing, I believe the link was in the article? http://poker.srv.ualberta.ca/
3x10^14 decisions isn't even in the same range of something like Chess or Go - you can tell this as we can't yet "prove" those games. A decent human can probably play a perfect game. Strange that poker champions tend not to be poker champions for long, unlike Chess champions, Go champions, etc.