The Baby Tartanian 8 could be to closest developers have come to creating the perfect no-limit hold'em heads-up poker bot after it won the latest edition of the Annual Computer Poker Competition in the Total Bankroll category and placed third in the Bankroll Instant Run-Off category. student at the School of Computer Science, designed Tartanian 8 alongside his adviser, Tuomas Sandholm, a professor in the Computer Science Department, specifically for the challenge of being able to play a strategy with the best results while using limited information.
Eleven teams competed for the title of best poker bot with the Carnegie Mellon University's Baby Tartanian 8 being awarded that honor by the Association for the Advancement of Artificial Intelligence. Tartanian 8 was developed as a successor to Tartanian 7 using the Comet supercomputer at the San Diego Supercomputer Center, and was built from scratch to help increase the speed that it processed algorithms, but had to be scaled down due to restrictions on the participants. "The algorithm is not specific to poker," explained Brown.
"It should be generally applicable to any strategic interaction where you have multiple agents and asymmetric information." Asymmetric information is where one person has information that the other person does not, such as your own hole cards in poker, or, to use a non-poker example, negotiation situations.
One major change to this latest poker bot was what Brown described as "pruning," which minimized the number of options that would be investigated while at the same time ensuring that the algorithm arrived at the optimal strategy.
"In a game like poker, where you have a lot of actions available to you, pruning helps cut down to actions that are worth investigating according to a given situation, instead of considering all the actions available to you," said Brown.
The Baby Tartanian poker bot was developed after the competition of the Brains vs Artificial Intelligence exhibition that ran from April 24 through to May 8, 2015.
During this time, the Claudico bot took on poker pros Bjorn Li, Doug Polk, Dong Kim, and Jason Les over 80,000 hands.
The humans finished the challenge $732,713 ahead (Li won $529,033, Polk $213,671, Kim $70,491, and Les lost $80,842), which seems like a wide margin, but more than $170 million was wagered during the challenge, meaning the humans' profit represented less than 0.5 percent of the total money wagered, which statistically made the challenge a tie.
Feedback from the four pros enabled Brown to make alterations to his new bot, changes he's continuing to make with the hope of one day being able to see his machine come out on top against the best poker players in the world.
"The ultimate goal in this line of research is to eventually beat the top humans," he said.
"Hopefully, in the next 2-3 years, we will be able to accomplish that." Duplicate Poker was used to determine the winner of this year's Annual Computer Poker Competition, around 15 million hands of it, in an attempt to remove much of the luck element.
Duplicate Poker is a variant of poker based on the principles of duplicate bridge where players play poker with a pre-set deck.