Author: Marc Lacrosse
Date: 07:35:47 11/15/05
Go up one level in this thread
On November 15, 2005 at 10:11:38, James Constance wrote: >Seems a bit fast to me. Do engines tend to perform relatively the same at >different time settings - I mean if it were played at 1 minute a move, would >GambitFruit do as well, or would it get hammered? Dear James, this question has been debated very often. In brief, there is a good (but surely not perfect) correlation between the relative strength or various engines independently of the timing used for testing. Although the correlation is imperfect, the advantage of testing at very small time limit is that you can compensate with a much larger number of games played, which is the only way to have some statistical significance in a short period when you are alone for the tests. So the question is not : is 1min+1sec a good time setting ? But well do I better evaluate the relative strength of two programs with 100 "1+1" games or with 10 "30 minutes KO" games. As 10 games produce no statistical signification at all, I prefer the 100 "very quick" games. But sure I would prefer 100 slow games ! Marc
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.