Author: Robert Hyatt
Date: 20:31:37 10/15/99
Go up one level in this thread
On October 15, 1999 at 15:38:44, odell hall wrote: >On October 15, 1999 at 15:25:03, Robert Hyatt wrote: > >>On October 15, 1999 at 15:08:59, odell hall wrote: >> >>>On October 15, 1999 at 15:00:11, Robert Hyatt wrote: >>> >>>>On October 15, 1999 at 13:25:17, Howard Exner wrote: >>>> >>>>>Anyone have the scoop on this re-match? Watching the one game on the rebel page >>>>>but how are the other three games unfolding? >>>> >>>> >>>>Rebel won 1 game, lost 2, and drew 1. >>> >>> >>> >>> Unfortunately a very disapointing result for Rebel, I bet your real happy! I >>>don't know what these two losses mean , since the I'ms are very strong and >>>capable of beating any grandmaster on any given day. >> >> >>_I_ happen to be pulling for the computer in every game. But I am realistic >>in my expectations of the outcome. This result wasn't bad. 1.5 vs 2.5 for >>4 IM players in 4 40/2hr games is not a bad result. It is right in line with >>what I would expect/hope for myself. 2-2 would have been very good. > > > > Well wouldn't Rebel's result be bad even if your projection was that Rebel was >a 2450 elo player? So I don't see how you could say this is not a bad result. >The problem is we are not using consistent hardware? What if I was to use a pent >133 and lost, would you factor in that result. Where do we draw the line as far >as counting the hardware? I wonder after all this grandmaster challenge is >over, what can we really say? Rebel is X strong on what hardware? since there >have been alot of differnet machines used. If I was to play rebel on four >diffenent machines How can I generalize as to what rebel strength is?? Since no >consistent machine is used? I hope you get what I am trying to say here. I >think this is another flaw in this experiment. Non consistency! All I would say is this: If that had been _my_ program, I would have been perfectly happy with that result at 40/2. Because I _know_ what humans are capable of, when it comes to exploiting weaknesses. The performance rating was around 2400, which is not bad. Especially when you consider that the "error" is at least +/- 100 with only those 4 games being considered. There are _many_ flaws. The version of Rebel is changing each game. The hardware changes. But those are 'noise' in the overall calculations. I doubt we could tell the difference between a K6/600 and a K6/400, when playing vs humans, at this extended time control...
This page took 0.01 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.