Author: leonid
Date: 18:59:13 09/13/99
Go up one level in this thread
On September 13, 1999 at 18:03:44, Bert van den Akker wrote: >For position: 2r1r1k1/3p1p1p/p2Qp1p1/q1N2P2/Pn1B1P2/6P1/2P3bP/2KRR3 b - - > >With Hash: > >Ply Time used Total time used >1 0 0 >2 0 0 >3 2 2 >4 4 6 >5 13 19 >6 9 28 >7 25 53 >8 115 168 > >Without Hash >Ply Time used Total time used >1 0 0 >2 0 0 >3 3 3 >4 3 6 >5 18 24 >6 34 58 >7 79 137 >8 324 461 > >For this tactical position the gain is almost 1 ply >because with hash a 8 ply search is done in 168 seconds >and without hash 7 ply search ply is done in 137 seconds. >The speed gain is (461/168)*100 =274% > > > > >Positional position: LCT 2 Test >position 11 >bn6/1q4n1/1p1p1kp1/2pPp1pp/1PP1P1P1/3N1P1P/4B1K1/2Q2N2 w - - 0 1 id "POS-11 > > >With Hash >Ply Time used Total time used >1 0 0 >2 0 0 >3 0 0 >4 1 1 >5 4 5 >6 4 9 >7 6 15 >8 19 34 >9 71 110 >10 288 398 > > >statistics for ply 10: >PV: h3-h4 g5xh4 f3-f4 h5xg4 f4xe5+ d6xe5 Pf1-h2 h4-h3+ Kg2-g3 >Pg7-h5+ Kg3-h4 Kf6-g7 Le2xg4 >Hash found/cutoff (6702436/308401) >Null move done/cutoff (101792/76889) >time = 236; score= 61 > >Without Hash >Ply Time used Total time used >1 0 0 >2 0 0 >3 0 0 >4 2 2 >5 2 4 >6 5 9 >7 11 20 >8 49 69 >9 149 218 >10 1588 1806 > >PV for ply 10: >h3-h4 g5xh4 f3-f4 h5xg4 f4xe5+ d6xe5 Pf1-h2 h4-h3+ Kg2-g3 Pg7-h5+ >Kg3-h4 Kf6-g7 Le2xg4 >time = 1393; score= 61 >Null move done/cutoff (361427/247503) > > >For this positional position the gain is more then 1 ply. >The speed gain is (1806/398) * 100 = 454% > >Conclusions for this positions: >For this position the gain is much more then 100% and I expect this will >grow for ply>10 if you hash table size is large enough. > > >If a position must be researched so as on ply 10 (positional position) where the >correct move h3-h4 was found hash tables will speed up the search because the >movesare well ordered by the hash table. > >I think hash tables are essential for move ordering. > >An other good thing to have for your move ordering is History tables. > >In addition the use of Recursive Null will speed up also the search with more >then 300% > > >Bert van den Akker Thank you very, very much for your numbers and explanations!!!!! Done just in a way that it is understandable and useful for people like me. Greate THANKS! Leonid.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.