Author: Robert Hyatt
Date: 18:33:45 11/11/00
Go up one level in this thread
On November 11, 2000 at 20:12:11, Bruce Moreland wrote: > >On November 11, 2000 at 18:33:52, Uri Blass wrote: > >>On November 11, 2000 at 18:20:25, Thorsten Czub wrote: >> >>>On November 11, 2000 at 14:36:02, Uri Blass wrote: >>>>I do not think that Hiarcs has big king attack scores relative to other >>>>programs. >>>> >>>>Uri >>> >>>wrong >> >>I think that your impression of bigger scores of hiarcs may be because of the >>fact that hiarcs has bigger scores in general than other programs. >> >>Here is an example that is not about king attack but about endgame: >> >>[D]8/2n4r/p1k1p3/PpPp1p2/1R1P1PpP/1P1R2P1/8/5K2 w - - 0 1 >> >>This position is from one of the games between Fritz6a and Nemeth Eduard >> >>Fritz6a evaluated this dead draw as +2.94 for white and Hiarcs7.32 can see +4.39 >>pawns for white. >> >>This is not because Fritz is smarter than Hiarcs but because of the fact that +2 >>of Hiarcs is not identical to +2 of Fritz. >> >>Uri > >I see this all the time in games with my engine vs Crafty. For example in some >drawn endings mine will say +2 and Crafty will say +1. > >Some engines contain a term that's supposed to induce them to trade down when >ahead material, I think. > >bruce The easiest way to implement a "trade-down" bonus is to set up an array which is used like this: score += trade[remaining_material]; trade[0] might be 200, trade[60] and above might be 0... which means that if you are a pawn up, with no pieces left at all, you might get +2 from the trade down bonus. I do this, and it tends to skew the scores. If material is even, this doesn't happen of course. I used to hate it, but now just tolerate it.
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.