Author: Dann Corbit
Date: 10:19:38 05/22/01
Go up one level in this thread
On May 21, 2001 at 22:32:55, Uri Blass wrote: >On May 21, 2001 at 21:20:40, Dann Corbit wrote: ><snipped> >>I don't think you can learn much of anything by looking at a single game. > >I agree that you cannot learn about 100 elo difference or 200 elo difference >from a single game but you still can learn more from one game then you learn >from the results. If the difference is 1000 ELO, it won't take long to find out. If it is 200 ELO, it will be difficult, but a couple hundred tests will decide the matter for certain. If the difference is 50 ELO, you will probably never know. >Here is an extreme example: >If you look at a game when one of the players play random moves you can learn at >a short time more than you learn from the results. > >Another example: >If you see that the first program is correct in pondering in 90% of the moves >when the second program is correct only in 70% of the cases then you learned >important information. Correct pondering can be a sign of weakness. What I mean is, one program may be impossibly stronger than the other. In the quiet moves, it always makes correct positional choices. Using its eval function, it expects the other program to make correct choices but it never does. So it ponders the opponent's move correctly only rarely. And yet this program is vastly superior. Now, in a strongly tactical position, there is (more or less) a 'correct' answer that most programs would agree on, given enough time. But these positions are a rarity. >This kind of information can lead you to guess that the first program is faster >than the second program and it helps it to ponder correctly and it increase the >probability that the first program is better. I still doubt the usefulness of examination of the individual moves of a game and trying to reach a conclusion from it.
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.