StrangeFish2 is a bot created to play Reconnaissance Blind Chess (RBC). It won the NeurIPS 2022 RBC tournament and placed second in the NeurIPS 2021 RBC tournament! For more information about RBC and the tournaments, see https://rbc.jhuapl.edu/ and https://rbc.jhuapl.edu/past_competitions.
StrangeFish2 is a continuation of the NeurIPS 2019 RBC tournament winner, StrangeFish. https://github.com/ginop/reconchess-strangefish. StrangeFish2 has a similar architecture, but is much more efficient compared to the previous implementation. More critically, the sense strategy has been modified and is much stronger.
StrangeFish is separated into two parts: the primary component manages game-state information throughout the game, and the secondary component makes decisions for sensing and moving based on that available information. Because of this separation, it should be simple for users to implement their own strategies to use within StrangeFish.
StrangeFish maintains a set of all possible board states by expanding each possible board into a new set for each possible opponent's move each turn. The set of possible board states is reduced by comparing hypothetical observations to in-game results for sensing, moving, and the opponent's move's capture square. Expansion and filtering of the board set are parallelized, and set expansion is begun during the opponent's turn (when possible).
The sensing strategy does calculate and consider the expected board set size reduction, but the primary influence on sensing decisions is the expected outcome of the following move choice. Before sensing, the bot calculates scores for each move option on each possible board (or for as large of a random sample as time allows). It also computes subsets of those boards which would be validated by each possible sense observation and estimates the likelihood of each observation. It then chooses moves for each of those possibilities and computes the expected result of that move based on the estimated probabilities of the boards in that subset. The sense choice is the square with the greatest expected outcome averaged over possible observations, move choices, and move results.
The sensing decision is therefore coupled with the move strategy. In this case, that strategy is a flexible combination of the best-case, worst-case, and expected (weighted-average) outcomes for each available move across the board set. The relative contributions of the mean, min, and max scores are tunable. The scores for each individual move were calculated using StockFish when possible, with a set of heuristics for evaluating board states unique to RBC such as staying or moving into check or actually capturing the opponent's king. The moving strategy also contributes to the board set size maintenance by awarding additional points to moves which are expected to eliminate possibilities (for example, a sliding move which may travel through potentially-occupied squares).
We use conda to manage our execution environment.
conda env create --file environment.yml
conda activate strangefish
export STOCKFISH_EXECUTABLE=/absolute/path/to/stockfish
export PYTHONPATH='.':$PYTHONPATH
See scripts/example_game.py
to play a local RBC game using StrangeFish2.
To connect to the server as was done for the NeurIPS 2021 tournament,
use scripts/modified_rc_connect.py, implemented here with
click command-line arguments.
Ex: python scripts/modified_rc_connect.py [your username] [your password] --max-concurrent-games 1