Author: Robert Hyatt
Date: 06:45:33 02/26/01
Go up one level in this thread
On February 26, 2001 at 03:51:01, Sune Larsson wrote: >On February 25, 2001 at 22:35:23, Robert Hyatt wrote: > >>On February 25, 2001 at 11:17:38, Sune Larsson wrote: >> >>> >>> [D]8/5b2/1k2p1p1/2NpP1P1/1K1P4/8/8/8 w - - 0 1 >>> >>> >>> In this position white's knight is far superior to black's bishop. This is >>> very old knowledge and white is winning this one. People that play French >>> as black might have nightmares of positions like this. The black bishop >>> can only defend the two weak pawns on e6,g6 or wander around like a ghost >>> on the white squares. >>> >>> These are the evals from some programs - PIII 800 - 3 min thinking time. >>> The higher +score the better. Remember, white has a won position. >>> And it's about evaluating a superior horse vs a bad, bad bishop. >>> >>> [1.Na4+ Kc6 2.Ka5 Kb7 3.Nc5+ Kc7 (3.-Ka7 4.Kb5 Be8+ 5.Kb4 Bf7 6.Ka5 Bg8 7.Kb5) >>> 4.Kb5 Be8+ 5.Ka6 Bf7 6.Nb7 Be8 7.Ka7 Bf7 8.Nd6 Bg8 9.Ka6 Kc6 10.Ka5 Bh7 >>> 11.Nf7 Bg8 12.Nh8 Bh7 13.Ka6 Kc7 14.Kb5 Kb7 15.Kc5 Kc7 16.Nf7 Bg8 17.Nd6 Bh7 >>> 18.Ne8+ Kd8 19.Nf6 and wins] >>> >>> >>> 1) AnMon 5.07 +1.05 >>> 2) Deep Fritz +1.00 >>> 3) Fritz 6 +0.97 >>> 3) Nimzo 7.32 +0.97 >>> 5) Nimzo 8 +0.90 >>> 6) Phalanx 22 +0.86 >>> 7) SOS +0.80 >>> 8) Crafty 18.01 +0.54 >>> 9) CM8K +0.51 >>> 10) Hiarcs 7.32 +0.48 >>> 11) Fritz 5.32 +0.47 >>> 12) Century 3 +0.43 >>> 13) Gandalf 4.32g +0.36 >>> 14) Junior 6 +0.32 >>> 15) Junior 5 +0.27 >>> 16) Gromit 3.1 +0.26 >>> 17) Chess Tiger +0.18 >>> >>> Sune >> >> >>Your score comparison is not really a good way to do this. You are >>assuming that +1.00 means the same for everybody. It doesn't. Some >>programs don't even count a pawn as +1.00, for example. A better way >>to compare is to take the same program and set up a position with a >>characteristic, and then without, and compare the scores of the same >>program for the two different positions, to see how much it values the >>characteristic you are testing... > > Yes, I guess you're right about this. Still, the above position is won > for white - the white knight can attack every single black pawn - the > black bishop cannot touch the white one's - the white king is just on the > edge of entering black's camp while his king is trying to hold on. > Black has no counterplay and it's possible to show a forced win, but, as I see > it, none of the programs really understands the danger black is facing. Though, > from this example, it looks like Deep Fritz has better knowledge about the > position than for example Fritz 5.32, Junior 5 and 6 or Tiger. The difference > in the evals here are 0.53, 0.73, 0.68 and 0.82 which is worth some notice, > I think - even if +1.00 doesn't mean exactly the same for everybody. > > Sune The point I was making was that in the case of Cray Blitz, an eval of +.400 was essentially winning, positionally. In the case of Crafty, +.40 is just a reasonably decent position in some cases, and almost winning in others. IE my scores are not _absolute_ in any sense (except for mates of course). What I consider most important is the 'positional' score (knowledge only) for the best move vs the next best move, to be sure that Crafty understands why the best move is really best. It would be nice if +x.xx meant _exactly_ +x.xx... but that is hard to do, which is why the programs all vary so much in terms of score, yet often choose the same move...
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.