Improving Sluggo through Hashing and Sub-branching 7.3 Adding Storage Techniques to the Hash Scheme to Make Sluggo Play Faster... 38 Appendix A: Induction Proof of (1-x) List of Figures List of Tables Title Improving Sluggo through Hashing and Sub-branching

Go, a board game which originated from China, is now very popular in east Asia and gaining its popularity throughout the rest of the world. Due to its large board size (19×19) and complex life-and-death analysis, computer Go is still one of the toughest artificial intelligence problems faced by computer scientists. Even though it has been studied for more than 20 years, computer Go programs still remain at a very low level compared to human players. The SlugGo group at the University of California at Santa Cruz has been working on this problem. Based on GNU Go, the best open source program, a full-board lookahead scheme which simulates the strategy of human players has been implemented to improve the performance of GNU Go. The enhanced GNU Go is called SlugGo. In order to get better performance, SlugGo analyzes several choices provided by GNU Go. To find the best choice from the candidates, SlugGo looks ahead, finding possible responding moves to these choices, until reaching a predetermined depth. By applying this lookahead scheme, SlugGo can play better than GNU Go. In this thesis, two techniques, hashing and sub-branching, are used to further enhance SlugGo. To reduce the time, the hashing technique stores the already calculated board positions and applies them to later calculations. An overall of about 10% timing performance gain is achieved. The sub-branching technique, rather than branching at the top only, branches to the maximal depth in the game tree allowed by the available computing resources. It is expected that sub-branching can make SlugGo play stronger. However, the preliminary results in this thesis show that further research is needed to improve this scheme.