Computers used to “solve” two-person, limit Texas hold’em
The article speaks about the progress computers have made when beating humans in certain games. These games that the computers have been so good at are called “perfect information game,” where both players know everything about the game. Whereas, “imperfect information games” are ones where players have info that their opponents don’t, such as card games. Researchers have focused on heads-up Texas hold ‘em, an example of imperfect information games, to see if they could figure out how to ensure a Nash equilibrium is reached.
Counterfactual regret minimization is the approach used in to help find that Nash equilibrium, which is when a losing strategy results in negative regret and winning strategy results in positive regret. The biggest problem with this is that it would take immense amount of space to store all the data, about 260 TB of disk space, but they successfully reduced that to about 17 TB of disk space. Strategies are chosen with likelihood proportional to the regret values. Researchers have gotten to a point where a person could play the computer for 70 straight years (12 hours a day) with their optimized strategy and still not have a large margin over the computer. The knowledge of game theory as a result of researching this game can lead to big findings with other issues.
This article relates to the concept of game theory that has been prevalent throughout the course thus far. With regards to heads-up Texas hold ‘em, the computer has to sort through the 10^161 circumstances that it could face and then decide on the best move based on the regret minimization approach. However, since the circumstances in the game change each time a player makes a move and the computer doesn’t have all of the opponent’s info, there can be multiple Nash equilibria. In the long run, despite however many games are played, neither the computer nor human will want to switch to a different strategy if they are at a Nash equilibria because it won’t be beneficial. This concept of applying game theory and Nash equilibria in such advanced technological situations goes to show that it can also be applied to solve larger-scale problems such as the economy.
Source: http://arstechnica.com/science/2015/01/computers-used-to-solve-two-person-limit-texas-holdem/