Advanced Algorithm Makes Poker a Solved Game
Oskari Tammelin, a Finnish software engineer, and a team of computer scientists from the University of Alberta in Edmonton, Canada, lead by Michael Bowling, have created an algorithm that can play poker perfectly. The algorithm plays a variant of poker called heads-up limit hold’em (texas hold’em with two players), so well, “as to render pointless further work on this game” (Eric Jackson, a computer-poker researcher). This means that the game is essentially solved, and the algorithm can win against any human player, no matter how skilled.
The algorithm uses game theory to predict whether an opponent is bluffing or not, and can even bluff itself. Poker is an example of a game with mixed strategy Nash Equilibrium, where two players do not have dominant strategies, and therefore must randomize their responses. While human players might fall into an equilibrium of bluffing and being truthful without conscience thought, the algorithm is able to record and reference all previous poker hands. This allows the algorithm to calculate with great accuracy the probability that an opponent is bluffing or not. The algorithm is also able to bluff the optimal amount, keeping its opponent from developing a dominant strategy. The algorithm is able to calculate the payoff matrix of poker hands to an extent that was not possible before.
The algorithm also uses advanced computer learning to adjust and learn from past experiences. It does this by assigning a “regret” value to each decision it makes, based on how good or bad the outcome of that decision was. The algorithm then uses what is essentially a network of past experiences to make smarter choices in the future.
Source: http://www.nature.com/news/game-theorists-crack-poker-1.16683
Very Nice, Thanks For this information.