Author: Robert Hyatt
Date: 07:41:04 05/31/01
Go up one level in this thread
On May 31, 2001 at 03:14:08, Robert Raese wrote: >On May 30, 2001 at 10:18:05, Robert Hyatt wrote: > >>On May 30, 2001 at 08:23:38, Ricardo Gibert wrote: >> >>>On May 30, 2001 at 08:07:19, Uri Blass wrote: >>> >>>>On May 30, 2001 at 07:05:02, Ricardo Gibert wrote: >>>> >>>>>He's a monster! Look what he did to scrappy (crafty) in an 11 game session on >>>>>ICC: >>>>> >>>>>82: + 3382 W 3365 scrappy [ br 5 7] A44 Res May 30 01 05:15 >>>>>81: = 3365 B 3382 scrappy [ br 5 7] A03 Agr May 30 01 04:57 >>>>>80: = 3364 W 3383 scrappy [ br 5 7] A43 Agr May 30 01 04:35 >>>>>79: = 3363 B 3384 scrappy [ br 5 7] B07 Agr May 30 01 04:10 >>>>>78: = 3362 W 3385 scrappy [ br 5 7] A40 Agr May 30 01 02:25 >>>>>77: = 3361 B 3386 scrappy [ br 5 7] A46 50 May 30 01 01:54 >>>>>76: = 3360 W 3387 scrappy [ br 5 7] D02 Rep May 30 01 01:39 >>>>>75: = 3359 W 3384 scrappy [ br 5 7] A40 Agr May 29 01 20:26 >>>>>74: + 3358 B 3385 scrappy [ br 5 7] B07 Res May 29 01 19:57 >>>>>73: = 3339 B 3414 scrappy [ br 5 7] B07 Agr May 29 01 18:11 >>>>>72: = 3335 W 3418 scrappy [ br 5 7] A40 Agr May 29 01 17:55 >>>>> >>>>>If he can do this at blitz, what chance would computer chess programs have >>>>>against him at slow chess? I think it is possible that Roman is the strongest >>>>>player in the world against computers. What does everybody else think about >>>>>that? >>>> >>>>I think that Crafty has the disadvantage that it is a free program. >>>> >>>>I guess that a lot of GM can beat Crafty at 5 7 time control in 11 games match >>>>if they train enough against it at home but other simply do not try it because >>>>they cannot earn much from it. >>>> >>>>I am more interested to know if JRLOK can do it against Ferret(In this case he >>>>has the problem that he cannot play against Ferret hundreds of games when Ferret >>>>and Bruce have no idea about the games). >>>> >>>>Uri >>> >>> >>>But it seems that Crafty has more "anti-anti-computer" code in it than other >>>programs, which are tuned to play other comps rather than humans. For example, a >>>lot of other programs succumb more easily to humans with simply a stonewall >>>attack, which does not work so well against crafty. >> >> >>This guy _is_ incredibly strong. But there are other things that are important >>here. IE he often wants to play vs a specific opening, which I always allow if >>asked. Several draws and a couple of wins is very unusual, but it can happen. >>By the same token I have seen him lose 20 straight too, of course... >> >>People are worried about Kramnik losing to deep fritz. This GM on ICC can >>find any small weakness in a program and drive a truck through it. And he will >>tell you what he found. And then it becomes amazing to see how once he finds >>such a weakness, he can then create the positions where that weakness suddenly >>becomes important. > >i've read that the main difference between grandmasters and average players is >that the "average" grandmaster has a cluster of about 40,000 position types he >can recognise like the back of his hand, and what ideas are important in them. >if true, this is an unbelievable advantage over a machine playing "blind" always >probing a few moves ahead. is there any way of approximating those "clusters" >in the machine, or am i just.... in way over my head again? > >:) Computers can do this with no problems. It is the "programmers" that are falling down here. :) IE evals are getting more and more special-case code, which is a direct correlation with the humans "patterns"...
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.