AI Strategy Approach Development on Baghchal using AlphaZero

Abstract

Baghchal, a traditional Nepali board game, involves strategic moves between tigers and goats on a 5x5 grid. The inscribed game board can be found in wooden structures or stone slabs in public places. Like every traditional game Baghchal is available in computer games. This paper implies the heuristic method and its enhancement with AI strategy. The heuristic method focused on evaluating tiger positions, movement patterns, and blocking strategies. In contrast, the Monto Carlo Tree Search (MTCS) approach used probability outcomes from multiple game simulations to optimize moves. The efficacy of heuristic patterns is derived from experienced players’ game-play and compared them with the computationally intensive approach of MCTS employed by AlphaZero. The findings indicate that while the heuristic approach can create a competitive bot, it hasl imitations in recognizing win/lose conditions early. The MCTS method provides a more robust strategy by averaging position evaluations over sub-trees. Through 20,000 simulated games, it is observed that tigers have an initial advantage, but optimal play can lead to draw. This dual approach offers comprehensive insights into Baghchal game-play, suggesting that combining heuristics with MCTS could enhance strategic decision-making in similar asymmetric board games.