SPECIAL THANKS:
https://github.com/psaikko/mcts-reversi
infrastucture and other algorithms AIs.
docker set up
docker build -t reversi .
docker run -it --rm reversi
running
cd /app/source
make
./run
Against 32*500 times MonteCarlo Simulator. (0.1s per turn)
| Opponent | Opponent's time per turn | Win Rate |
|---|---|---|
| random_move | 0s | 100% |
| greedy move | 0s | 100% |
| generous move | 0s | 100% |
| greedy move with 10 sampling | 0.01s | 55% |
| greedy move with 100 sampling | 0.07s | 25% |
| greedy move with 1000 sampling | 0.5s | 15% |
| UCT with 10 sampling | 0.004s | 60% |
| UCT with 100 sampling | 0.06s | 15% |
| UCT with 1000 sampling | 0.5s | 5% |
| UCB1 with 10 sampling | 0.004s | 55% |
| UCB1 with 100 sampling | 0.05s | 15% |
| UCB1 with 1000 sampling | 0.5s | 15% |
| MinMax eval_sampling_10 3 | 0.1s | 20% |
| MinMax eval_pieces 3 | 0.002s | 100% |
| MinMax eval_pieces 4 | 0.02s | 0% |
| MinMax eval_pieces 5 | 0.1s | 100% |