Author: Vincent Diepeveen
Date: 05:23:08 01/08/01
Go up one level in this thread
On January 08, 2001 at 08:16:44, Severi Salminen wrote: >Hi! > >This position from a Kramnik-Leko game was posted recently here by Eduard >Nemeth. I tried it with my program and found odd behaviour. (See below) >[D]2br1r2/1pR1Rp2/1p3k2/1B1p1Pp1/PP1P2P1/K7/8/8 w - - > >This is the output of my engine if null moves are disabled (you won't see >end-of-ply statistics except at the end of search. A info line is printed when >PV changes): >-------------------------- > Time Depth Score Nodes PV > 0.00 1 0.39 7 a5 > 0.00 1 0.53 9 Bf1 > 0.00 1 0.59 11 Be2 > 0.00 1 0.63 13 Bd3 > 0.00 1 0.66 25 Re5 > 0.00 2 0.53 68 Re5 Rh8 > 0.00 3 0.58 233 Re5 Rh8 Kb3 > 0.00 3 0.60 1065 Kb2 Rd6 Re5 > 0.00 3 0.61 1256 Kb3 Rd6 Re5 > 0.06 4 0.58 1694 Kb3 Rd6 Be8 Rd8 > 0.17 5 0.68 7665 Kb3 Kg7 Kc3 Rd6 Re5 > 0.44 6 0.56 28541 Kb3 Kg7 Re5 Rh8 Bd3 Rd6 > 2.37 7 0.66 149581 Kb3 Kg7 Kc2 Kf6 Kd3 Rd6 Re5 > 11.10 8 1.16 672797 Kb3 > (++) > 14.72 8 1.29 858649 Kb3 Kg7 Bd7 Bxd7 Rxd7 Rxd7 Rxd7 b5 > 44.88 9 1.29 2557418 Kb3 Kg7 Bd7 Bxd7 Rxd7 Rxd7 Rxd7 b5 Rxd5 > >Eval: 1.29, nodes: 3007933, qnodes: 4116797, time: 117.60, nps: 81456 >Futile moves: 12023153, order: 99.0% >------------ > >And this is with null moves: >----------- > Time Depth Score Nodes PV > 0.00 1 0.39 7 a5 > 0.00 1 0.53 9 Bf1 > 0.00 1 0.59 11 Be2 > 0.00 1 0.63 13 Bd3 > 0.00 1 0.66 25 Re5 > 0.00 2 0.53 69 Re5 Rh8 > 0.00 3 0.58 277 Re5 Rh8 Kb3 > 0.05 3 0.60 1149 Kb2 Rd6 Re5 > 0.05 3 0.61 1358 Kb3 Rd6 Re5 > 0.05 4 0.48 1662 Kb3 Rd6 Re5 Rh8 > 0.22 5 0.65 7311 Kb3 Rd6 Kc2 Kg7 Re5 > 0.38 6 0.58 18869 Kb3 Rd6 Re3 Rd8 Re5 Rh8 > 0.99 7 0.66 65854 Kb3 Kg7 Kc2 Kf6 Kd3 Rd6 Re5 > 2.19 8 1.16 156746 Kb3 > (++) > 2.80 8 1.29 210619 Kb3 Kg7 Bd7 Bxd7 Rxd7 Rxd7 Rxd7 b5 > (--) > 5.65 9 1.29 456674 Kb3 Kg7 Bd7 Bxd7 Rxd7 Rxd7 Rxd7 b5 Rxd5 > >Eval: 1.29, nodes: 199941, qnodes: 301868, time: 6.15, nps: 81594 >Futile moves: 498682, order: 94.7% >------- > >So both versions fail high at d=8 trying Kb3, but why does the null move version >fail low at d=9 but research with infinite window results the same score? The >PVs (and scores) are the same so I think everything is working right. Is this >kind of behaviour "normal" when using null moves or do I have a bug? Can >aspiration search and null moves cause these articats? i see you have 'futile moves' also. If you throw futility pruning out, then nullmove should work brilliant positionally spoken. Otherwise you reduce depth and let opponent to move then you prune without trying the reduced depth even! Usually that's tactically not missing too much if you implement a few exceptions like allowing all captures, but it prunes positionally a lot! Nullmove definitely should give also in endgame about same PV and same score. Nullmove never can cause big problems, it can at most show existing problems. Greetings, Vincent >Severi
This page took 0.01 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.