Author: Dann Corbit
Date: 16:08:07 08/09/00
Go up one level in this thread
On August 09, 2000 at 16:59:40, Mark Young wrote: [snip] >>>1...Nf3+ >>> ± (1.03) depth: 12/36 00:00:52 38577kN >>>1...Nf3+ 2.gxf3 Qh4 3.Rh1 Bxh3 4.c3 Rf6 5.Bxf4 Rxf4 6.Qe3 >>> = (0.25) depth: 12/36 00:00:53 39290kN >>>1...Nf3+ 2.gxf3 Qh4 3.Rh1 Bxh3 4.c3 Rf6 5.Bxf4 Rxf4 6.Qe3 >>> = (0.00) depth: 13/32 00:00:55 40746kN >>>1...Nf3+ 2.gxf3 Qh4 3.Rh1 Bxh3 4.c3 Rf6 5.Bxf4 Rxf4 6.Qe3 >>> = (0.00) depth: 14/38 00:01:04 47886kN >>>1...Nf3+! >>> = (-0.16) depth: 15/35 00:01:37 73480kN >>>1...Nf3+! 2.gxf3 Qh4 3.Rh1 Bxh3 4.Bxf4 Rxf4 5.Qe3 Rf6 6.Qd4 >>> µ (-0.84) depth: 15/39 00:02:08 97818kN >>> >>>(Young, 09.08.2000) >> >>It chose that move rapidly, but the evaluation shows that it is more a random >>glitch than a computer brilliancy. When it gets the eval right, then I will >>believe that the computer has made a brilliant move. It sees an advantage of >>less than one pawn, so it does not know how good the move is yet *or* why it >>should choose that one. It is rather impressive that the computer got to 15 >>plies in a minute and a half, and a hundred million nodes in two minutes though. >> Must be a pretty nice machine. > >It looks like it knows what its doing to me, what is the correct line of play >here, that is more telling then the eval score. BTW Fritz 6a eval keeps climbing >as it searches deeper and it sticks with Nf3+. I think it is a leap of logic to >suggest that fritz 6a does not understand that Nf3+ is not only the best more >here, but it is also winning. I think you will be hard pressed to call a 3.00 >score a random glitch. By the time the eval has risen to a full piece, it is very clear that the program knows it is winning. At ply 15, it sees its advantage as less than one pawn. By the time it has finished ply 17, it seems to know what is going on. An advantage of less than one pawn (to me) means that the program does not know it is winning. Even with one pawn up or down, outcome is extremely uncertain. However, by the time you are up 3 pawns in a deep eval, computers are almost never going to lose against any opponent.
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.