Solving Imperfect Information Games Using Decomposition
Authors: Neil Burch, Michael Johanson, Michael Bowling
AAAI 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | All results were generated on a 2.67GHz Intel Xeon X5650 based machine running Linux. To demonstrate the practicality of re-solving subgame strategies, we started with an almost exact Nash equilibrium (exploitable by less than 2.5 10 11 chips per hand), computed the counterfactual values of every hand in each subgame for both players, and discarded the strategy in all subgames. Figure 3 shows the exploitability when using a different number of CFR iterations to solve the re-solving games. In Figure 5, we demonstrate re-solving subgames with a Leduc Hold em strategy generated using an abstraction. |
| Researcher Affiliation | Academia | Neil Burch, Michael Johanson and Michael Bowling Computing Science Department, University of Alberta {nburch,johanson,mbowling}@ualberta.ca |
| Pseudocode | No | The paper describes algorithms but does not provide structured pseudocode or an algorithm block. |
| Open Source Code | No | The paper does not provide any statement or link indicating that the source code for the described methodology is publicly available. |
| Open Datasets | Yes | To show that re-solving subgames introduces at most an arbitrarily small exploitability, we use the game of Leduc Hold em poker, a popular research testbed for imperfect information games (Waugh et al. 2009; Ganzfried, Sandholm, and Waugh 2011). |
| Dataset Splits | No | The paper describes the game and how it was split into a trunk and subgames for the decomposition approach but does not provide explicit training, validation, or test dataset splits (e.g., percentages or sample counts) typically used for reproducibility in machine learning. |
| Hardware Specification | Yes | All results were generated on a 2.67GHz Intel Xeon X5650 based machine running Linux. |
| Software Dependencies | No | The paper mentions using CFR variants (CFR-D, PCS) and an operating system (Linux) but does not provide specific version numbers for any software libraries, frameworks, or tools used to implement the experiments. |
| Experiment Setup | Yes | Our implementation of CFR-D used CFR for both solving subgames while learning the trunk strategy and the subgame re-solving games. All the reported results use 200,000 iterations for each of the re-solving subgames (0.8 seconds per subgame.) Each line of Figure 4 plots the exploitability for different numbers of subgame iterations performed during CFR-D, ranging from 100 to 12,800 iterations. There are results for 500, 2,000, 8,000, and 32,000 trunk iterations. |