It seems my transposition table works perfectly fine, but the numbers on saved number of nodes is different than what I expected. I'd like to know if there's an answer to this, or that it depends on the engine. I have implemented a transposition table, and also sort on the hash move. These are the results in the KiwiPete position:
Hash table in use, but with both hash move ordering and hash cuts disabled (the table always returns "nothing found" for this test).
Code: Select all
info score cp 25 depth 1 seldepth 16 time 0 nodes 4271 nps 0 pv e2a6 b4c3 b2c3 e6d5
info score cp 25 depth 2 seldepth 16 time 0 nodes 8576 nps 0 pv e2a6 b4c3 b2c3 e6d5
info score cp 20 depth 3 seldepth 20 time 2 nodes 23332 nps 11666000 pv e2a6 e6d5 a6b7 b4c3 b7a8 c3d2
info score cp 20 depth 4 seldepth 20 time 7 nodes 62976 nps 8996571 pv e2a6 e6d5 a6b7 b4c3 b7a8 c3d2
info score cp 5 depth 5 seldepth 22 time 28 nodes 257843 nps 9208679 hashfull 1 pv e2a6 b4c3 d2c3 e6d5 e4d5 f6d5
info score cp 15 depth 6 seldepth 22 time 140 nodes 1116882 nps 7977729 hashfull 16 pv e2a6 e6d5 c3d5 b6d5 a6b7 a8d8 b7d5 f6d5
info score cp 5 depth 7 seldepth 24 time 679 nodes 5635001 nps 8298971 hashfull 62 pv e2a6 e6d5 c3d5 f6d5 e5c4 f7f5 c4b6 a7b6
info score cp -15 depth 8 seldepth 26 time 4006 nodes 28315376 nps 7068242 hashfull 371 pv e2a6 e6d5 c3d5 f6d5 e5d3 f7f5 g2h3 f5e4
info score cp -40 depth 9 seldepth 27 time 23292 nodes 167732837 nps 7201307 hashfull 923 pv d5e6 e7e6 e2a6 e6e5 g2h3 b4c3 d2c3 e5e6 c3f6 g7f6
Code: Select all
info score cp 25 depth 1 seldepth 16 time 0 nodes 4271 nps 0 pv e2a6 b4c3 b2c3 e6d5
info score cp 25 depth 2 seldepth 16 time 1 nodes 8576 nps 8576000 pv e2a6 b4c3 b2c3 e6d5
info score cp 20 depth 3 seldepth 20 time 2 nodes 23332 nps 11666000 pv e2a6 e6d5 a6b7 b4c3 b7a8 c3d2
info score cp 20 depth 4 seldepth 20 time 6 nodes 60738 nps 10123000 pv e2a6 e6d5 a6b7 b4c3 b7a8 c3d2
info score cp 5 depth 5 seldepth 20 time 28 nodes 255524 nps 9125857 hashfull 1 pv e2a6 b4c3 d2c3 e6d5 e4d5 f6d5
info score cp 15 depth 6 seldepth 22 time 138 nodes 1090083 nps 7899152 hashfull 16 pv e2a6 e6d5 c3d5 b6d5 a6b7 a8d8 b7d5 f6d5
info score cp 5 depth 7 seldepth 24 time 643 nodes 5377310 nps 8362846 hashfull 59 pv e2a6 e6d5 c3d5 f6d5 e5c4 f7f5 c4b6 a7b6
info score cp -15 depth 8 seldepth 26 time 3854 nodes 27042610 nps 7016764 hashfull 359 pv e2a6 e6d5 c3d5 f6d5 e5d3 f7f5 g2h3 f5e4
info score cp -40 depth 9 seldepth 27 time 22048 nodes 158657343 nps 7195997 hashfull 909 pv d5e6 e7e6 e2a6 e6e5 g2h3 b4c3 d2c3 e5e6 c3f6 g7f6
Code: Select all
info score cp 25 depth 1 seldepth 16 time 0 nodes 4271 nps 0 pv e2a6 b4c3 b2c3 e6d5
info score cp 25 depth 2 seldepth 16 time 1 nodes 8576 nps 8576000 pv e2a6 b4c3 b2c3 e6d5
info score cp 20 depth 3 seldepth 20 time 2 nodes 23332 nps 11666000 pv e2a6 e6d5 a6b7 b4c3 b7a8 c3d2
info score cp 20 depth 4 seldepth 20 time 6 nodes 60618 nps 10103000 pv e2a6 e6d5 a6b7 b4c3 b7a8 c3d2
info score cp 5 depth 5 seldepth 20 time 26 nodes 231149 nps 8890346 hashfull 1 pv e2a6 b4c3 d2c3 e6d5 e4d5 f6d5
info score cp 15 depth 6 seldepth 22 time 108 nodes 849402 nps 7864833 hashfull 12 pv e2a6 e6d5 c3d5 b6d5 a6b7 a8d8 b7d5 f6d5
info score cp 5 depth 7 seldepth 24 time 425 nodes 3347752 nps 7877064 hashfull 40 pv e2a6 e6d5 c3d5 f6d5 e5c4 f7f5 c4b6 a7b6
info score cp -15 depth 8 seldepth 26 time 1923 nodes 13492089 nps 7016167 hashfull 248 pv e2a6 e6d5 c3d5 f6d5 e5d3 f7f5 g2h3 f5e4
info score cp -40 depth 9 seldepth 27 time 9306 nodes 64753214 nps 6958222 hashfull 817 pv d5e6 e7e6 e2a6 e6e5 g2h3 b4c3 d2c3 e5e6 c3f6 g7f6
In my situation, the hash cuts seem to provide a much bigger saving than the hash move ordering.
Going by the descriptions on some of the websites, I was expecting world-shattering savings on the hash move sorting, and marginal savings on the hash cuts, but it seems to be the other way around. Do you have any thoughts about why this could be?