Author: Dieter Buerssner
Date: 14:18:28 11/20/03
Go up one level in this thread
On November 20, 2003 at 17:09:12, Russell Reagan wrote: >"For instance, some experiments found that improving 5 times the running time of >such an algorithm the playing strength of a chess programs improves about 200 >Elo points." [...] >That would mean a 2x speedup was worth 80 ELO. No. You must think in a logarithmical fashion here, not in a linear fashion. You seem to have calculated 80*5/2. So 100 time more time would give 80*100/2. That seems unreasonable. Typically, it is assumed (without any real convincing argument behind it, however) that each doubling is worth about the same amount. The more reasonable formula would calculate 86 Elo (each dobuling means more or less by definition that Elo increase is proportional to log of time, not to time itself). log(5)/log(2)*86 = 200. Regards, Dieter
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.