Conclusion

Extended futility pruning and limited razoring apply the futility idea to pre- and pre-pre-frontier nodes. Slight modifications and additions with respect to normal futility pruning provide the proper safety precautions that greatly reduce the selective risks of the new schemes. Extensive experiments with 2180 positions from the well-known test suites ECM, WAC, and WCS provide quantitative empirical evidence for the benefits of each scheme alone and of both combined. While scaling well with search depth and hardly compromising any tactical strength, extended futility pruning and limited razoring reduce the search effort of our strong and tournament-proven chess program DARKTHOUGHT by additional 10%-30% on average at fixed search depths of 8-12 plies. Moreover, extended futility pruning demonstrated its practical viability with a successful real-life performance under tough tournament conditions during the 15th WMCC in October 1997. These convincing results show that statically selective forward-pruning techniques do not only promise some success but really deliver on making alpha-beta search more scalable at high search depths.



Created by Ernst A. Heinz, Thu Dec 16 23:28:11 EST 1999