| Scoby the Poker Robot | |||||
| The Challenge of Poker Bots by Lars Sundin
There is a lot of discussion going on about poker bots, that is, computer programs that can play poker. Some fear that the online poker boom may come to a crashing halt if tables are swarmed by robots, mechanically grinding away and taking all the money off the unaware humans and completely unmoved by the fuming outbursts in the table chat. Others claim they’d be happy to play against robots every day of the week, suggesting that poker robots will not be able to put up a very strong opposition. Towards a Deep Blue of poker Where the truth lies remains to be seen, as poker bots is a rather young and unexplored domain. In the history of chess playing computers, there was a time when the discussion stood at pretty much the same position. Chess computers up till then weren’t very strong. International champions and other competent human opponents had no problem beating the computers, and many were happy with the assertion that the human brain couldn’t be outsmarted by machines. We all know the current position in that discussion. Chess computers play even matches against world champions, and most other players stand a snowball’s chance in hell against a strong computer. The ability of computers to calculate extreme numbers of moves, coupled with an ever more intricate position analysis, has overcome the human brain which can see patterns and concentrate on essentials to an extent unmatched by computers, but which lacks the flawless memory capacity of silicon chips. Why wouldn’t poker bots go the same way? Well, it’s hard to prove, but there are some valid arguments against this. First of all, unlike chess, poker is a game of incomplete information. Not even massively parallel super computers can perform exact calculations from imprecise input. Dealing with unclear premises and fuzzy fragments of information comes much more natural for the human brain which is used to it from its everyday operation – no offense. Probability calculations at the speed of light The solution for a computer in this case, of course, is to go at the business in a fuzzy manner. Instead of calculating exact moves and half-moves as in chess, the computer will resort to estimating probabilities for various events, based on the actual information at hand and educated speculations about what distribution of hands opponents might hold depending on their betting patterns. Here, the computer is back at its home turf and can carry out billions of correct probability calculations in the blink of a diode. Nevertheless, the result can never be better than the original educated speculation, which again is not the strong point of the rigid computer “brain”. Secondly, and probably more important, poker is not a game of cards and chips alone. It is also a game of people. The correct decision in most situations depends heavily on the personalities of your opponents and their current frame of mind. Much more so in no-limit than fixed-limit games, by the way. In theory, to take a fully informed decision in no-limit poker you would need to know everything that has ever happened to your opponents while reading their set of genes off their DNA cards. Even if this is a bit exaggerated, at least your decision making must be based on the opponents’ past and recent behavior. This is no easy task for a human player, as anyone knows who has ever played a hand of poker. Of course, computers can keep full records of all your earlier actions at the table, again overtrumping the brain by its stringency. But understanding and interpreting records of human behavior is a fuzzy task more suited for the human brain than for a computer. | |||||
| Top Players | News | Directory | Forum | FAQ | How to Win | Store | Contact | Home | |||||
|
| ||||
|
| ||||
