Uploaded September 2026 | Updated September 2026, 1 hour ago
JassBotAI explores how artificial intelligence can make the traditional Swiss card game Schieber-Jass playable even when four people are not available.
The mobile app, developed in Unity for iPhone and iPad, allows users to play complete Schieber games either offline against AI opponents or online with any combination of friends and computer-controlled players. Online multiplayer is implemented using Unity Relay, while free seats can automatically be filled by bots.
Instead of relying on predefined rules, the AI opponents are based on specialized neural networks for trump selection and the different game modes. Thousands of games are generated through self-play, and the resulting data is used to train the models in Python with PyTorch. The trained models are then exported to ONNX and integrated directly into Unity.
The project demonstrates how machine learning can be combined with a traditional Swiss card game to create flexible and intelligent digital opponents while preserving multiplayer gameplay with friends.
This work was developed as part of a Bachelor Thesis at Bern University of Applied Sciences (BFH).
JassBotAI explores how artificial intelligence can make the traditional Swiss card game Schieber-Jass playable even when four people are not available.
The mobile app, developed in Unity for iPhone and iPad, allows users to play complete Schieber games either offline against AI opponents or online with any combination of friends and computer-controlled players. Online multiplayer is implemented using Unity Relay, while free seats can automatically be filled by bots.
Instead of relying on predefined rules, the AI opponents are based on specialized neural networks for trump selection and the different game modes. Thousands of games are generated through self-play, and the resulting data is used to train the models in Python with PyTorch. The trained models are then exported to ONNX and integrated directly into Unity.
The project demonstrates how machine learning can be combined with a traditional Swiss card game to create flexible and intelligent digital opponents while preserving multiplayer gameplay with friends.
This work was developed as part of a Bachelor Thesis at Bern University of Applied Sciences (BFH).
