Author: Roy Eassa
Date: 09:37:51 10/14/03
Go up one level in this thread
On October 14, 2003 at 05:05:20, Daniel Clausen wrote: >On October 13, 2003 at 17:39:03, Russell Reagan wrote: > >[snip] > >>G/5 34 points >>G/10 24 points >>G/30 15 points >>G/120 23 points >> >>Only the last result increases. If you ran this test numerous times with many >>different engines, I bet you'd see the trend of closer results with longer time >>controls. > >Based on the four numbers stated above I'd _never_ conclude that the number is >decreasing when time increases. > >Sargon The margin decreased in two of the three time increases. Once by 10 points and once by 9 points. The one time it increased was by 8 points. Way too little data to pronounce a real trend, I agree, but it does seem to have some trend-like characteristics IMHO. :-) Combine it with the many other results that have been posted and I think a trend is indeed showing itself. Longer time controls = narrower margins of victory. Not always, but usually. Working backwards from that apparent trend, shorter time controls = wider margin of victory AND a lot more games can be played. So shorter time controls are generally a better way to compare 2 engines, for those two reasons. (Admittedly, the "value" of each game to a human who's seeking near-perfect chess is much lower, but that's an entirely separate issue.)
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.