Research Center for Digital Sustainability

Can the Best Jass AIs Beat the Top Humans?

This project is available for both Bachelor's and Master's students.

Introduction

Despite the recent successful application of Artificial Intelligence (AI) to games, the performance of cooperative agents in imperfect information games is still far from surpassing humans. Cooperating with teammates whose play-styles are not previously known poses additional challenges to current state-of-the-art algorithms. In the Swiss card game Jass, coordination between the two opposing teams is crucial for winning. Since verbal communication is forbidden, the only way to transmit information within the team is through a player’s play-style. This makes the game a particularly suitable candidate subject to continue the research on AI in cooperation games with hidden information. The best computer programs can compete with strong amateur players in the Schieber variant of the popular Swiss card game Jass [1]. Convolutional neural networks (CNN) are on par with Monte Carlo Tree Search (MCTS) in Schieber.

Research Questions

Experiments have shown that the best Jass AIs are able to compete with good amateur players. So far, they have not played against top human players.

RQ1: How do the best Jass AIs compete against top human players?

Steps

Find top human Schieber players and motivate them to join the experiment
Set up the experiment environment
Conduct the experiment
Evaluate the results

Activities

⬤⬤⬤◯◯ Programming

⬤⬤⬤⬤◯ Experimentation

⬤◯◯◯◯ Literature

Prerequisites

Good programming skills (preferably in Python)

Experience with Machine Learning and Neural Networks

Contact

Joel Niklaus

References

[1] Niklaus, J., Alberti, M., Ingold, R., Stolze, M., & Koller, T. (2020). Challenging Human Supremacy: Evaluating Monte Carlo Tree Search and Deep Learning for the Trick Taking Card Game Jass. ICAISC.