Act1 (2016)

Act1 was the 3rd place agent in the 2016 AAAI Computer Poker Competition instant runoff event.

Act1 was trained by an experimental distributed implementation of the Pure CFR algorithm. A heuristic was added to occasionally avoid some game tree paths, reducing the time spent per training iteration. To compensate for imperfect recall, a distance metric that considers features from all postflop streets was used to construct the card abstraction on the river. Several bet sizes were omitted because they offer little benefit against other equilibrium opponents while requiring a disproportionate amount of resources to train and store. (This turned out to be a risky gamble that had mixed results depending on the opponent.)

The strategy consists of 159 billion information sets (430 billion information set-action pairs) and completed 5.15 trillion iterations. It was trained for ~180000 CPU hours and is 15x larger than its predecessor Prelude, whom it defeats by 4 bb/100.

See this post on the 2016 ACPC results.