Author: Bas Hamstra
Date: 16:41:39 08/17/99
Go up one level in this thread
Bob, Double NPS <hijg> ? Could you explain lazy eval a bit more? Doing only material eval to get Beta cutoffs or something? What about big positional scores? I saw Ed Schroeder describe it as very dangerous, once...? Regards, Bas Hamstra. On August 16, 1999 at 15:50:00, Robert Hyatt wrote: >On August 16, 1999 at 07:54:28, Phil Dixon wrote: > >> >> >> >> >>Halwick Jr,F (2430) - Dikmen,A (2207) [B12] >>corr M.040 IECG, 1998 >> >>1.e4 c6 2.d4 d5 3.e5 Bf5 4.Nc3 e6 5.g4 Bg6 6.Nge2 c5 7.h4 h6 8.Be3 Qb6 9.f4 Nc6 >>10.f5 Bh7 11.Qd2 0-0-0 12.0-0-0 c4 13.Nf4 Qa6 14.fxe6 Nb4 15.exf7 Ne7 16.g5 Bxc2 >>17.Bh3+ Kb8 18.Rde1 Bh7 19.a3 Ng6 20.gxh6 gxh6 21.e6 Nxh4 22.Qf2 Be7 23.Qg3 Ka8 >>24.Kd1 Nd3 25.Nfxd5 Rxd5 26.Nxd5 Nxb2+ 27.Ke2 c3+ 28.Kf2 Nd3+ 29.Kg1 Qd6 30.Qxd6 >>Bxd6 31.Rf1 1-0 >> >>While looking at this game with Fritz, I noticed the node speed hit 500k on move >>9... and 1,500k on move 10... and 1,200k on move 11. Qd2. Only very briefly, >>1-2 seconds and at low ply levels. I am wondering if this is a rather common >>occurrence at low plys or if it has something to do with the position? >> >>I am running a 233 Mhz with 64 Megs of RAM. > > >The best possible node speed comes when move ordering is bad. IE Fritz seems >to depend on the hash table for this, and at the start of a new position search, >it has no idea... (IE I doubt it uses history ordering, etc, although I have no >idea). As the hash table gets useful stuff in it, the nps will drop. > >Another place is in wild positions where 'lazy evaluation' becomes very >effective (ie in most variations lots of material has been won or lost, >which makes lazy eval take almost zero time and can easily double the nps >[at least in crafty]) > >Bob
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.