Author: Howard Exner
Date: 19:53:01 07/23/98
Go up one level in this thread
On July 23, 1998 at 12:56:12, Howard Exner wrote: >On July 23, 1998 at 05:12:45, odell hall wrote: > >>Are there any longer doubts that micro's have reached the Grandmaster Level? I >>hope the the so-called computer experts will not insult our intelligence and >>still deny that micros have reached this level. If not then please show me one >>international master in the world that could achieve the same results against >>annand in a match or for that matter grandmaster? > >Bob Hyatt's post on Deep Thought made lots of sense a while back. >He said that they took 25 consecutive games played against humans >at standard time controls and rated it. This is what we humans do for ourselves >so what's good for the goose ... > >In Rebel's case we should just tally up its last 25 games against humans >and gets its performance rating. Ditto for other programs that play against >humans. > >Ed could do this. The events that come to mind are Anand (2),Dominican >Republic Tournament(5) and three Aegon events(18) ,so that is 25 games. >The time controls at Aegon were slightly slower so some may argue the >method and say the rating should be made a bit lower. But on the flip >side one could argue that the rating could be increased a bit to factor >in the slower hardware from three years ago. I think this applys >to Junior also, which has already a nice performance rating vs humans >at standard time despite older results on slower machines. > >Is this "cherry picking" (ie:just choosing programs that have done well >vs humans and disregarding the rest)? Not at all because each program >since created by different individuals should have a unique status and >stand on their own. Take ICC for example. Ferret has a whopping 3000 >(or something near that) rating (this of course at blitz which now >around 8min/game average due to the popularity of the increment >time controls). No one is suggesting that because >Ferret has this that all programs should have it. Its Ferret's rating >alone after thousands of games and no one elses. We often clump all >the programs together but perhaps we should just follow suit and do >as we treat ourselves in chess. > >The different programmers no doubt have a record of their performance >ratings vs humans. They can share them ,as some already have in the past, >if they so choose. So Rebel does have a rating based on 25 consecutive games >of WXYZ as accurate as any human's rating after 25 games. Here is a summary of Rebel's achievement vs humns at long time controls. AEGON 95: TPR 2473 (Rebel7) P-90 Game 90'/30" 6 games AEGON 96: TPR 2530 (Rebel8) P-166 Game 90'/30" 6 games AEGON 97: TPR 2619 (Rebel9) P-200 Game 90'/30" 6 games A 40 move game would be 110 minutes per side without the pressure of a flag falling which would definitely aid the human since the computer doesn't understand pressure. I hear it likes to inflict it though :-) Are these the machines used for these different years? Just going on my memory for machine speed. Dominican Republic Rebel 9 played 5 Fide Rated opponents at 40/2 then all/1 hr. on a P-225 Eddy De Los Santos (2190) 1-0 Jose M. Dominguez (FM 2305) .5-.5 Gustavo Hernandez (2320) 1-0 Nelson Pinal (IM 2330) 1-0 Ramon Mateo (IM 2415) 1-0 TPR =Rc+400(W-L)/N TPR=2312+400(4)/5 TPR=2632 Note: this tournament was 11 rounds. Rebel defeated Hiarcs and drew CM5000 in this tournament. It also defeated 4 other participants who do not have a fide rating, for a first place score of 10/11. Rebel 10(AMD-450) v Anand (.5-1.5) TPR =2795+400(-1)/2 TPR = 2595 Is Rebel's play against humans in the GM range?
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.