Author: Eugene Nalimov
Date: 14:03:07 01/16/03
Go up one level in this thread
Simplest probable explanation is: you are probing everywhere when you have the opportunity. Crafty (and I presume some other programs) probe only first M plies. E.g. when they are doing N-ply search from the root position they probe only first N*2 plies (or N*3/2, or ... -- you got the idea). Thanks, Eugene On January 16, 2003 at 16:33:52, David Rasmussen wrote: >On January 16, 2003 at 15:54:17, Dann Corbit wrote: > >>> >>>Beowulf output: >>>Position 1 - White to play >>>-------------------------- >>>ID = >>>FEN = 8/4n3/1k6/1p4p1/1P4P1/1K6/3B4/8 w - - >>> >>>. . . . . . . . >>>. . . . n . . . >>>. k . . . . . . >>>. p . . . . p . >>>. P . . . . P . >>>. K . . . . . . >>>. . . B . . . . >>>. . . . . . . . >>> >>>Current Position = 0.52 >>>Number of Possible Moves = 11 >>> >>> 11 236 75 143184 Bxg5 Ng6 Kc3 Ne5 Be3+ Kc6 g5 Kd7 Kd4 Ke6 g6 >>> 12 231 178 352812 Bxg5 Ng6 Kc3 Ne5 Be3+ Kc6 g5 Kd5 Bd4 Nf3 g6 Kd6 >>> 13 219 266 528614 Bxg5 Ng6 Kc3 Ne5 Bd8+ Kc6 g5 Kd7 Bf6 Ng4 Kd3 Ke6 >>>Bd8 Kf5 >>> 14 213 606 1216211 Bxg5 Nd5 Bd2 Kc6 g5 Ne7 Kc3 Kd5 Kd3 Ke5 Be1 >>> 15 213 1009 1942140 Bxg5 Nd5 Bd2 Kc6 g5 Ne7 Bc3 Kd5 Bf6 Ng6 Kc2 Ke4 >>>Bd8 Ke5 Bf6+ Ke4 &rep. >>> 16 214 1759 3325204 Bxg5 Nd5 Bc1 Kc6 Kc2 Nxb4+ Kb3 Nd3 Bd2 Kd5 Kc3 >>>Ne5 Kb3 >>> 17 214 3000 5733741 Bxg5 Nd5 Bc1 Kc6 g5 Ne7 Kc2 Kd5 Kd3 Ke5 Bd2 Nd5 >>>Be1 Ne7 >>> >>>Search Time : 30.00 Seconds >>>Total Nodes Searched : 5733741 (52.3% Qui) >>>Total Nodes Evaluated : 2998609 >>>Cuts : Delta 1508 : SEE 18270 : Mate 0 >>> : Razor 202628 : Eval 1107291 >>>Extensions : Check 151560 : OneRep 45 : CMThreat 8317 >>> : Pawn 42183 : Recap 2342 : RevCheck 4278 >>>Hash Stores (Size) : 582291 (699037) >>>Hash Probes (Hits) : 5384740 (3263270) >>>EGTB Probes (Hits) : 46423 (46423) >>>Move Ordering : 81.9% >>>Best Move = Bxg5 >>>Move Score = 214 >>>--Answer: Bxg5 ** Correct ** >>>Time To Solution = 0.00 sec >>>Correct So Far : 1/1 >>> >>> >>>Test Completed. >>>Results; >>>Correct : 1/1 >>>Average Time To Solution = 0.00 sec >>>[1]W > >> >>Same thing with EGTB turned off: >> >>Position 1 - White to play >>-------------------------- >>ID = >>FEN = 8/4n3/1k6/1p4p1/1P4P1/1K6/3B4/8 w - - >> >>. . . . . . . . >>. . . . n . . . >>. k . . . . . . >>. p . . . . p . >>. P . . . . P . >>. K . . . . . . >>. . . B . . . . >>. . . . . . . . >> >>Current Position = 0.52 >>Number of Possible Moves = 11 >> >> 11 236 68 145925 Bxg5 Ng6 Kc3 Ne5 Be3+ Kc6 g5 Kd7 Kd4 Ke6 g6 >> 12 219 106 229472 Bxg5 Ng6 Kc3 Ne5 Bd8+ Kc6 g5 Kd7 Bf6 Ke6 Bd8 Ng6 >>Kd4 >> 13 218 178 396062 Bxg5 Ng6 Kc3 Ne5 Bd8+ Kc6 g5 Kd7 Bb6 Ke6 Kd4 Nc4 >>Bd8 Nd6 >> 14 211 336 771335 Bxg5 Ng6 Kc3 Ne5 Bd8+ Kc6 g5 Kd5 Bf6 Ng6 Kd3 >>Ne5+ Kc2 Ke6 Kd2 Nf7 Ke3 >> 15 211 621 1421370 Bxg5 Ng6 Kc3 Ne5 Bd8+ Kc6 g5 Kd5 Bf6 Ng6 Kd3 >>Ne5+ Kc2 Ke6 Kd2 Nf7 Ke3 >> 16 207 1947 4503141 Bxg5 Nd5 Bd2 Kc6 g5 Ne7 Kc2 Kd5 Kc3 Ke5 Kb3 Nc6 >>Be3 Kf5 Bf4 >> 17 207 3000 6998206 Bxg5 Nd5 Bd2 Kc6 g5 Kd7 Kc2 Ke6 Kd3 Kf5 Kd4 Nxb4 >>Kc5 Na2 Kxb5 Nc3+ Kc4 Ne4 >> >>Search Time : 30.00 Seconds >>Total Nodes Searched : 6998206 (53.9% Qui) >>Total Nodes Evaluated : 3765466 >>Cuts : Delta 1878 : SEE 24636 : Mate 0 >> : Razor 292648 : Eval 1481897 >>Extensions : Check 181048 : OneRep 96 : CMThreat 0 >> : Pawn 71075 : Recap 3354 : RevCheck 7484 >>Hash Stores (Size) : 742253 (699037) >>Hash Probes (Hits) : 6474752 (3664648) >>EGTB Probes (Hits) : 0 (0) >>Move Ordering : 81.9% >>Best Move = Bxg5 >>Move Score = 207 >>--Answer: Bxg5 ** Correct ** >>Time To Solution = 0.00 sec >>Correct So Far : 1/1 >> >> >>Test Completed. >>Results; >>Correct : 1/1 >>Average Time To Solution = 0.00 sec > >This seems to indicate that >1. Beowulf probes actually more in the position. ~45000 probes vs. ~25000 >probes. >2. Beowulf loses a little speed when using EGTBs, but not nearly enough as >Chezzz. Beowulf goes from 700M/30s to 600M/30s. Chezzz goes from 700knps to 100 >knps. What hardware is Beowulf running on here, and how much hash, egtb cache >etc. ? > >/David
This page took 0.02 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.