Scripts of Tribute AI Bot

This repository contains our AI bot developed for the Scripts of Tribute AI competition, a re-adapted version of the card game Tales of Tribute from The Elder Scrolls Online.

---

The game

Scripts of Tribute is a two-player, competitive card game where each player builds and adapts their deck to outplay the opponent.
Players manage three main resources:

• Coins: for purchasing cards
• Power & Prestige: for winning through score advantage
• Patron Favor : for alternative win conditions

A match ends when:

1. A player reaches 40 prestige and maintains the lead until hitting 80 prestige, or
2. A player secures the favor of all patrons.

From an AI perspective, the game presents a large and complex state space with stochastic elements, strategic phases (early, mid, and late game), and the need for adaptive decision-making.

---

Our Approach

We designed an AI bot based on Monte Carlo Tree Search (MCTS) enhanced with heuristics:

• Prior move and Prior Choice heuristic : to reduce the search space and improve performance.

• MCTS: Simulates possible moves within a player’s turn.

• Heuristic Evaluation: A custom metric called Min-Max Hand Value Rating (MMHVR) evaluates game states by considering:

  • Resources (prestige, power, coins, patron favor)
  • Deck composition and quality
  • Strategic potential across game phases

• Evolutionary Optimization: Heuristics were refined automatically through evolutionary algorithms.

Results: Our agent outperformed both a random bot and a simpler baseline strategy, demonstrating the effectiveness of MMHVR within MCTS.

---

Usage

Install required packages:

pip install -r requirements.txt

Run a match between two bots:

python3 main.py

Project Structure

• BotCommon/: Shared codebase (including heuristic implementation)
• bots/: Implementations of the different bots
• ExampleBot/: example bots provided by the documentation
• HeuristicLearning/: Evolutionary algorithm for heuristic optimization
• MCTS/: Implementations of the varius version or MCTS algorithm
• helpers/: Helper functions for logging and game management

---

Contact

• Vito Barra – v.barra1@studenti.unipi.it
• Alessio Iacullo – a.iacullo@studenti.unipi.it