Rion Snow

Here we present the algorithms and extensions that produced our successful Othello-playing program, which finished among the top five programs of this year’s Othello tournament. Specifically, we discuss a variety of search algorithms and extensions, as well as our algorithms for adaptive timemanagement, our efficient ‘bit board’ data structure, our features comprising our evaluation function, our two learning methods, our opening book, and our endgame solver. Throughout the report we discuss the experiments and evaluations that led us to our particular design decisions, including comparison of the efficiencies of search strategies, performance analysis of our data structures, and testing of different feature sets. 1 Search Algorithms and Extensions We began our study of search algorithms with the implementation of standard Minimax search with alpha-beta pruning. We quickly discovered that our initial implementation of search was far too slow to win the tournament, which inspired us to explore a variety of means by which to reduce the number of nodes we search and thereby improve the speed of our algorithm. Most important of these were implementations of a transposition table, an efficient move-ordering scheme, and explorations of the advanced search algorithms NegaScout and MTD(f).