Author: stuart taylor
Date: 04:37:43 11/07/03
Go up one level in this thread
On November 06, 2003 at 22:32:18, Terry McCracken wrote: >On November 06, 2003 at 22:02:49, Steve Maughan wrote: > >>Terry, >> >>>SSDF is NOT a scientific measuring stick, it's not meant to be. >> >>It's the closest we have to scientific testing in Computer Chess. >> >>>Actually a 100 ELO difference by a margin of 50> or 50< is nearly meaningless >>>with comuters, and I dare say with humans as well. >> >>Hmmm, after having answered you other post I now realize you trolling.... >> >>You clearly know little about statistics > >Sir, I'm not trolling at all, and I know what I'm talking about. > >Do you remember Alan Tomalty? Well I know him through his work and and have >talked with him, he'd agree with me. > >Also, I've tested computers myself and have played in rated tournaments, I >understand what I'm talking about. > >So you musn't have a very good grasp what a rating means. It's _not_ a >scientific or absolute measuring of strength. It is statistical only and if >you understand statistics, you'll understand my meaning of of 50 points up or >down have little meaning. > >However, I know you'll disagree. I think it's best if te _quality_ of the play can also be evaluated. I don't normally get satisfactory replies when I ask about this. So the nearest hope of estimation seems to me to be the ssdf. If, e.g. a program plays extremely riskily, and entertainingly, but gets 50 elo below first place, then I have serious thoughts that that program might be the greatest, and if it was configured to be more careful, at the expense of other qualities, It might well get a higher elo than the top one of now. i.e, its wins are much more worthy (with its 50 elo deficit to top) and intersting and instructive, and its losses are more petty in meaning. 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.