Author: blass uri
Date: 03:00:07 06/01/99
Go up one level in this thread
On May 31, 1999 at 18:00:29, Robert Hyatt wrote: >On May 30, 1999 at 10:04:41, blass uri wrote: > >>More than 70% of the voters say that deep blue chip will not have more than 2800 >>ssdf rating. >> >>In other words top programs on pentium200 can expect more than 20% against it. >> >>I abstained in the opinion poll because I do not know if to believe Hsu's >>results of 38:2 of weaker version(deep blue Junior) or to believe my impression >>based on public games that deep blue Junior is not so strong. >> >>Deep blue Junior lost against 24xx player in 15 minutes per game and drew >>against another not strong player. >> >>Deep blue Junior won most of the games but every commercial could do the same. >> >>Uri > > >Let me tell you a short 'story' to explain why such 'polls' are useless. >Several years ago, I posted a query in r.g.c (before it was split) about >Cray Blitz and a 'sparring opponent' asking for advice on which program or >dedicated machine to buy to use as a test opponent when tuning Cray Blitz >for a tournament (ACM). Most everyone said 'get genius... you won't have >a chance against it, the commercial programs are so much stronger than your >program, because all you have is horsepower, not 'algorithms'. The argument >went on for a good while. > >I got genius, ran it on the 486/66 (fastest PC available at the time) and >played several matches against Cray Blitz. I started at 5 minute games, but >after Cray Blitz was 10-0 ahead, I gave up. I then reduced Cray Blitz to 1 >cpu (rather than 16) which effectively slowed it by a factor of 12. It >won 20 in a row. I then gave genius 30 minutes per game, cray blitz 5. It >won the next 10. Now 40 in a row, no draws or losses. All games were absolute >tactical busts. I finally got closer when I I gave genius a 30:1 time handicap >and only let CB use 1 cpu to boot. Overall a 30*12 : 1 handicap. And at that >handicap, out of the next 10 games, CB won 7, drew 1 and actually lost 1. > >The "poll" happened because _nobody_ knew anything about Cray Blitz. They >"assumed" (by listening to others that also didn't know, but who weren't afraid >to act like they did) that the CB guys didn't know what we were doing. It >became pretty apparent that we _did_. > >Does this story sound familiar? IE flash forward to DB and today? It tore >hell out of the best commercial programs at a 100:1 time handicap. Yet no >one thought it would have a chance even slowed down to 'equal' speed. The >problem is that no one is listening to Hsu. No one wants to listen to Hsu. >Everyone wants to believe that their precious microcomputers are 'close' to >DB. The propblem is that people do not believe Hsu My impression based on the games of DB against kasparov without additional information is that it is better than the commercials but not so good that top programs of today cannot get even 10% against it. My impression based on public games of Deep blue Junior that I saw is that it is not better than the commercial programs. My impression may be wrong but if the 38:2 of Deep blue Junior against top commercial programs is right then I think that he should refuse to play against humans with deep blue Junior after the match against kasparov and agree to play only against the commercial programs. Uri
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.