Hadron Game AI

This repository contains the AI logic for Hadron, a strategic game-playing AI that dynamically combines two powerful decision-making algorithms: Alpha-Beta Pruning and Monte Carlo Tree Search (MCTS).

---

🧠 AI Strategy Overview

The core function powering Hadron's decision-making is:

def montecarlo_alphabeta_search(game, state):
    if len(game.actions(state)) > 7:
        return monte_carlo_tree_search(game, state, 800)
    return h_alphabeta_search(game, state)

🔍 How It Works

• When the number of legal actions is greater than 7, the AI assumes the state is too complex for deterministic methods and uses Monte Carlo Tree Search (MCTS) to sample possible outcomes.
• When the number of legal actions is 7 or fewer, it switches to a more precise Heuristic Alpha-Beta Pruning algorithm.

This hybrid approach allows the AI to scale well in complexity while retaining accuracy in simpler states.

---

🧩 Search Algorithms

🔺 Alpha-Beta Pruning (alphabeta_search, h_alphabeta_search)

• A classic optimization of the minimax algorithm.
• Efficiently prunes the game tree to avoid exploring unnecessary branches.
• h_alphabeta_search adds unusual heuristic functions—odd, non-intuitive evaluations that, despite seeming strange, have shown to work surprisingly well for Hadron’s game mechanics.
• The heuristics used aren't conventional, matches with an odd number of actions available are preferred, but they consistently improve decision-making performance based on extensive testing.

🎲 Monte Carlo Tree Search (monte_carlo_tree_search)

• Best suited for large and unpredictable decision spaces.
• Simulates random playouts from possible actions to estimate long-term value.
• The implementation here uses 800 simulations per move for a balance between performance and precision.

---

💡 Why This Hybrid Works

• MCTS brings strength in wide, unpredictable scenarios where exploration is key.
• Alpha-Beta shines in narrow, tactical situations where deeper evaluation matters.
• By combining both and tuning when each is used, Hadron’s AI can adapt fluidly between early-game chaos and late-game precision.

---