Author: martin fierz
Date: 08:55:23 02/12/04
Go up one level in this thread
On February 12, 2004 at 10:57:50, Djordje Vidanovic wrote: >On February 12, 2004 at 06:03:37, Uri Blass wrote: > >>The main point is that programs should not evaluate unclear position as winning. >> >>Black has a big advantage without capturing the pawn so there is no reason to >>capture h3 and get unclear pawn endgame. >> >>evaluating Rxh3 Rh2 as a win for white without search is a msitake but >>evaluating it as a win for black is also a mistake that cause Crafty to miss a >>simple win by not capturing the pawn. >> >>Uri > >I disagree: you could evaluate it as a win for black after capturing the h3 >pawn if you have very precise evaluation. Of course, I am talking about Martin >Fierz's modification. This just means that Crafty has a faulty eval for this >type of sharp and unclear endgame pawn races. To illustrate that it is possible >to have a good evaluation take a look at how the new strong program, Patriot >(Alpha stage) does: > >New game >8/7p/8/5p1k/4p3/r6P/P5R1/6K1 b - - 0 1 > >Analysis by Patriot v2.0: > >1...Rxh3 2.Rg7 h6 3.Rf7 Kg4 4.a4 Ra3 5.Rg7+ Kf3 6.Rf7 f4 > -+ (-4.93) Depth: 10 00:00:00 448kN, tb=24 [snip] hi djordje, you cheated :-) you are already much too deep in the search at 10 ply plus extensions to show what patriot does in it's eval. what you really have to do is take the position after ...Rxh3 Rh2 Rxh2 Kxh2 and show what it says there - like this output of my program: you can see that on the first 3 ply, my eval thinks it's totally winning for black, but then the score drops again around ply 5, and jumps back up again. this shows that my eval sometimes awards a big bonus for black, and sometimes not (which comes from evaluating the pawn race at the start, obviously not too well...). anyway, if you look at very low search depth, you will see what the static eval of the program thinks about the outcome of the pawn ending - which is what i would like to see. give it this position (winning for black) and the one with the pawn on e5 (winning for white) and check when it sees what. for example, my thing first evaluates the position with the pawn on e5 as a black win at depth one, then it gets kind of agnostic, and only at depth 14 does it notice that it's white who is winning. which is kind of disappointing ;-) cheers martin black pawn on e4 Muse 0885 UCI 97 MB: D1+ 00:00 4.96 Kh5g4 D1 00:00 4.97 Kh5g5 D2+ 00:00 4.61 e4e3 D2 00:00 4.61 e4e3 Kh2g1 D3 00:00 4.51 e4e3 Kh2g1 Kh5g5 D4 00:00 1.72 e4e3 Kh2g1 e3e2 Kg1f2 e2e1Q Kf2xe1 D5 00:00 1.01 e4e3 Kh2g1 h7h6 a2a4 e3e2 D6+ 00:00 4.73 Kh5g5 D6 00:00 4.73 Kh5g5 Kh2g1 Kg5f4 a2a4 Kf4e5 a4a5 D7+ 00:00 4.63 Kh5g4 D7 00:00 4.63 Kh5g4 a2a4 Kg4f3 Kh2g1 f5f4 a4a5 e4e3 D8 00:00 3.67 Kh5g4 a2a4 Kg4f3 a4a5 e4e3 a5a6 e3e2 a6a7 e2e1Q a7a8Q Qe1e4 D9 00:00 3.70 Kh5g4 a2a4 Kg4f3 a4a5 e4e3 a5a6 e3e2 a6a7 e2e1Q a7a8Q Qe1e4 Qa8a3 Qe4e3 D10 00:00 2.99 Kh5g4 Kh2g2 Kg4f4 a2a4 Kf4e5 a4a5 Ke5d5 a5a6 Kd5c6 a6a7 D11 00:00 3.93 Kh5g4 a2a4 e4e3 Kh2g1 Kg4f3 Kg1f1 Kf3e4 a4a5 Ke4d5 a5a6 Kd5c6 D12 00:01 3.67 Kh5g4 a2a4 Kg4f3 a4a5 e4e3 a5a6 e3e2 a6a7 e2e1Q a7a8Q Qe1e4 Qa8a3 Qe4e3 Qa3a8 Kf3f4 D13 00:02 3.87 Kh5g4 a2a4 h7h5 a4a5 e4e3 Kh2g1 Kg4f3 Kg1f1 h5h4 a5a6 e3e2 Kf1e1 D14 00:05 9.03 Kh5g4 a2a4 e4e3 Kh2g2 f5f4 Kg2f1 Kg4f3 Kf1e1 Kf3g2 Ke1e2 f4f3 Ke2xe3 f3f2 Ke3d4 f2f1Q with pawn on e5 Muse 0885 UCI 97 MB: D1+ 00:00 4.31 Kh5g4 D1 00:00 4.32 Kh5g5 D2 00:00 -0.42 h7h6 a2a4 D3+ 00:00 -0.03 e5e4 D3 00:00 -0.03 e5e4 a2a4 e4e3 D4 00:00 -0.37 e5e4 a2a4 h7h6 a4a5 D5 00:00 0.92 e5e4 a2a4 e4e3 Kh2g1 e3e2 D6 00:00 -0.16 f5f4 Kh2g2 e5e4 a2a4 f4f3 Kg2f2 h7h6 D7+ 00:00 0.52 Kh5g4 D7 00:00 0.57 Kh5g4 a2a4 e5e4 Kh2g2 e4e3 Kg2f1 f5f4 D8+ 00:00 -0.20 Kh5g4 D8 00:00 -0.20 Kh5g4 a2a4 e5e4 Kh2g2 f5f4 Kg2f2 e4e3 Kf2e2 h7h5 D9 00:00 0.26 Kh5g4 a2a4 e5e4 Kh2g2 e4e3 a4a5 h7h6 a5a6 e3e2 D10 00:00 0.00 Kh5g4 a2a4 e5e4 Kh2g2 e4e3 a4a5 f5f4 Kg2f1 f4f3 a5a6 D11 00:00 -0.03 Kh5g4 a2a4 e5e4 Kh2g2 e4e3 a4a5 f5f4 a5a6 h7h5 Kg2f1 Kg4f3 D12 00:00 0.04 Kh5g4 a2a4 e5e4 Kh2g2 e4e3 a4a5 f5f4 a5a6 f4f3 Kg2f1 Kg4f4 Kf1e1 h7h5 D13 00:01 -0.03 Kh5g4 a2a4 e5e4 a4a5 e4e3 Kh2g1 f5f4 a5a6 Kg4g3 Kg1f1 Kg3f3 Kf1e1 D14+ 00:02 -4.40 e5e4 D14 00:03 -4.40 e5e4 a2a4 e4e3 Kh2g2 Kh5g6 a4a5 e3e2 Kg2f2 h7h5 Kf2e1 h5h4 a5a6 h4h3 a6a7 h3h2 D15 00:03 -4.32 e5e4 a2a4 e4e3 Kh2g2 Kh5g6 a4a5 h7h5 a5a6 h5h4 a6a7 h4h3 Kg2f1 f5f4 a7a8Q Kg6g5 D16 00:06 -4.64 e5e4 a2a4 e4e3 Kh2g2 Kh5g4 a4a5 f5f4 a5a6 f4f3 Kg2f1 h7h5 a6a7 Kg4g3 a7a8Q e3e2 D17 00:07 -4.66 e5e4 a2a4 e4e3 Kh2g2 Kh5g4 a4a5 f5f4 a5a6 f4f3 Kg2f1 h7h5 a6a7 Kg4g3 a7a8Q e3e2
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.