Author: odell hall
Date: 11:53:12 10/16/99
Go up one level in this thread
On October 16, 1999 at 14:38:56, odell hall wrote: >On October 16, 1999 at 04:26:10, Ratko V Tomic wrote: > >>> I don't see how anyone can call a result of 4-4 against 2400+ players >>> a bad result, under normal time controls. >> >>I guess, if someone takes the SSDF or similar comp-comp ratings as the real >>rating interchangeable with human play ratings, one then expects a 2600 SSDF >>program to convincingly beat 2450 player pool (70% points). In fact, if one were >>to continue playing against the same human team, week after week, even if >>Rebel's book gets updated and engine parameters adjusted between the matches, >>its performance rating would keep dropping (not necessarily uniformly, but as a >>moving average over a longer sequence). And if the fixed Rebel were to play such >>sequence, the performance rating would drop much faster. >> >>As to which level it would eventually stabilize at -- it would be the rating >>just a bit above the strength of the weakest aspect of the program (probably >>would settle somewhere between 2200 and 2300 on a 400-500 Mhz Pentium, due to >>lack of planning beyond its 12-15 plies). Blitz, of course, is a different ball >>game. (But so is the postal chess, in which none of the micro programs would >>come out above the 2200.) > > > > First of all I don't know of anyone that is claiming programs are 2600+ at >40/2. The projections I have heard was 2500+. And since Rebel is very near that >rating now, I don't see why this is considered so rediculous. You state that you >"believe" the rating would drop if the match was repeated seven times. Can you >give some statistics or evidence to back this up? Frankly the idea that humans >somehow are going to improve 100 pts against humans over time is somewhat >suspect to me. Could you give me a specific example of what you mean? Are you >saying the human for instance will discover some weakness in how the program >handles the kings indian, or sicilian and thus exploit that weakness? I guess >this whole idea is unclear. Since the computers book will navigate it around >many of these problems I really cannot see what you mean. Something else I think is worth considering is the Fact that We already have atleast 12 games at 40/2 from rebel, yet I cannot notice any drastic decline in rebel 's elo as the Grandmaster challenge continues. If your assertion was correct we should see such a decline. However rebel's elo has been pretty stable. Obviously humans are not learning anything that is important enough to make a difference. Another thing that is interesting, Did you read the rebel homepage? The grandmaster that Drew Rebel in the last lithuainian challenge explained that "honestly" he did not expect to beat rebel. I don't know if this comment was lost on everyonelse, but I certainly took notice. When a grandmaster makes such a statement it is natural to start thinking of the reasons why? Obviously he has the program and has evaluated it. If he as a grandmaster did not think he could beat it, I think that says mountains about the potential of the program.
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.