Author: Bruce Moreland
Date: 09:17:25 12/01/00
Go up one level in this thread
On December 01, 2000 at 09:16:12, martin fierz wrote: >i was wondering about some evaluations which fritz 5.00 gives me for the >following position: > >[D]1k1r3r/p1p1n3/1pq1p2p/P3Ppp1/2QP4/8/2P1NPPP/RR4K1 w - - 0 0 > >the black king is under attack, but fritz gives about +1.5 on depth 11 >for qxc6 (to be fair, this is only the second-best move according to fritz). >after you play out the moves qxc6 nxc6 the evaluation drops immediately to 0, >which is obviously a better evaluation for the position after the queen >exchange. >is this an example for preprocessing in action - bad black king AND queens on >the board? and if yes: which programs do NOT make this type of mistake? has >anybody who has a program which does a lot of preprocessing ever tested his >preprocessing version against a version which does no preprocessing? to me this >example with the queen exchange and an evaluation which jumps by 1.5 looks very >suspicious, but of course the program gets faster - does this compensate the >loss in accuracy of the evaluation? does this compensation change when the >search goes deeper? > >cheers > martin I don't know how preprocessors work, but my program won't play Qxc6, and I bet a lot of others won't either. It was very excited about Qa6 for a while, with an advantage of almost two pawns, but now it's settled down to +1.3, based upon the attack. If you want to find other positions where queens go off stupidly, you might be able to find a program that plays Qh3xQf1+ rather than Qh5 in the main line Marshall Gambit Ruy Lopez. Mine used to do that, and I'm pretty sure it was not the only one at the time. bruce
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.