Tuesday, September 18, 2012

Perft for Checkers for Depth 26

I made some improvements to my distributed implementation for computing perft and divide numbers for 8x8 checkers, and now computed depth 26 from the initial position, shown below. As reported earlier, the numbers were computed on a cluster of machines, further optimized with a "hard collision"-free transposition table as well as bulk counting. The move generator does not eliminate duplicate captures (viz. the situation where a king can capture the same pieces in different directions; a situation that starts to occur at depth 12 and up).

move              divide(26)
----------------------------
12-16:    111362678435231752
11-16:    112590257768420515
11-15:    101352649993886926
10-15:    100552646996749293
10-14:     86312674861234785
 9-14:    103811787278058952
 9-13:    136189763354484914
----------------------------
perft(26) 752172458688067137