Author: stuart taylor
Date: 02:46:03 10/11/02
Go up one level in this thread
On October 10, 2002 at 02:56:58, Daniel Clausen wrote: >On October 09, 2002 at 21:17:45, stuart taylor wrote: > >[snip] > >>And you anyway need about 500 games to come to a sensible estimation. > >There are several ways to to find out whether X is better than Y. One of them is >to play countless games and simply see who gets more points. This is typically >done in all the engine-engine matches and/or what the SSDF does. > >Another way is to play a few games but _objectively_ analyze the games >afterwards. This way doesn't work very well when comparing 2 engines, because >they style of play is similar, in a bigger sense. It works quite good in the >Kramnik-DF match though, in my opinion. > >While it's very well possible, that the match turns in the next games, I have no >doubt, that Kramnik indeed is stronger than DF, because I think the way he won >his games so far are so convincing. (it's not that DF had bad luck or something, >it simply lost chanceless - in my opinion) > >I wish people wouldn't be so focused on the match score of a match with only 8 >games. It really doesn't say a whole lot, whether it's 4.5 - 3.5 or 4-4 at the >end. But people are so focused on these numbers. I suggest that we enjoy the >games rather than counting beans. :) > >Sargon Thanks for your answer. So you say that numerous games is only necesary with Comp-comp games. But not when there is at least one human. I never saw someone say that, but it seemed to stare me in the face that no one was bothered by that question of what about humans? I was thinking it might have just been a question of virtual impossibility to get humans to play so many games the whole time, and I'm still wondering if that sticky point is making some of us biassed in our judgement as to what's necesary for evidence of playing standard and comparrisons. It's well known that many humans have those that they beat, and those that they lose to, irrespective of general rating.(which is a problem) AND I'm not yet convinced that there can not be found a way of analysing level from a sample of comp-comp games. S.Taylor
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.