Author: scott farrell
Date: 06:33:11 01/01/03
Go up one level in this thread
On January 01, 2003 at 08:48:39, Omid David Tabibi wrote: I think if assessing positional evaluations functions, it might be better to limit the search depth, say just try between 1 and 5, so the engine doesnt bring tactics in to try to force the material balance, or castle sooner, or some such thing. If you control the search depth, us humans can quickly see if there are any tactics at all in say a 3 ply search. And thus its the positional evaluator and not the search/null moves/etc etc that is the larger influencer. My engine with a static eval scored it at exactly 1 centipawn advantage for black (0.01) . After a few plies it got up to 0.5. At ply 8 it was back to 0.01 again. I think the static eval, or say ply 1-3 would be more useful for analysis. Scott >[D] r1bqk1nr/pppp1ppp/2n5/2b1p3/1PB1P3/5N2/P1PP1PPP/RNBQK2R b KQkq b3 0 4 > >The Evans Gambit, arising after 1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.b4, is a good >example of positional material sacrifice. By sacrificing his b pawn, white gains >a number if precious tempos, thus seizing the opening initiative. While the 4.b4 >gambit might not be better than the other 4th moves here, it usually reaches a >balanced position in which white has enough compensation for the sacrificed >pawn. > >Now let us see how the top engines evaluate this position. There is not much >tactics involved here, so this gives us a good opportunity to compare the >programs' evaluation (their chess knowledge). > >In the table below, the evaluation of each engine is recorded after 1 minute >analysis (since the evaluation is largely positional, no significant score >change was noticed from one ply to another, so most probably even deeper >searches will not change the result): > > >Engine Score >------ ----- >Junior 7 0.06 >Fritz 7 -0.41 >Shredder 6.02 -0.33 >Chess Tiger 14 -0.82 (Gambit Tiger 2 also produced the same score) >Hiarcs 8 -0.84 >Crafty 19.1 -0.81 > > >Interesting points: > > - Junior 7 was the only program who evaluated the position realistically. > > - Gambit Tiger 2 which is tuned for sacrificial play, did not evaluate > the position differently from Chess Tiger 14. > > - Hiarcs 8 which is said to incorporate the most chess knowledge, displays > the least chess understanding in this position! (Conclusion: more chess > knowledge does not necessarily mean better evaluation.) > >Omid. > >P.S. An analysis of Fritz 8 and Shredder 7 will be appreciated.
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.