Machine learning training with a Neural Network: Comparing trained robots in Schnapsen

Aslihan Neva Can, Emnet Tesfay, Teodora Gabriela Socola

Abstract: The study aims to iteratively train a Schnapsen bot and observe whether the win rate is improving against RDeep compared to its previous generation. “The history of the interaction of machine learning and computer game-playing goes back to the earliest days of Artificial Intelligence” [10]. However the amount of research papers since are quite limited. Only a few papers have attempted to go deeper into card games using machine learning. This study intends to make that gap smaller. The methodology relies on various tournaments held between all the generations of the Schnapsen bot and against RDeep, a more advanced bot. The results reveal that Iterations do indeed have an effect on the win rate of the game bot and do improve it significantly. The implications that this paper holds is for the broader field of machine learning and the application of card games. Also it adds to the limited studies on Schnapsen.