Fitting AlphaZero model with some tweaks to the chess variant game pawn wars
Trying both: Incentivized Self Play and Games Vs. Stockfish
- White has been induced to play more king moves (2000 games)
- Black is pure self play (2000 games)
- Built upon White model above then 8000 selfplay games (10000 games)
- Built upon White model above: then 8000 selfplay games (10000 games)
The model seems to employ the Fox in the Chicken Coop strategy