Author: Dieter Buerssner
Date: 10:13:23 07/31/05
Go up one level in this thread
On July 30, 2005 at 05:40:03, Sune Fischer wrote: >You can try and detect the lag by comparing the time you had at move n - minus >the time you spent searching, with the time you have at move n+1. >These should match to +-0.005 second, but usually time has mysteriously been >lost somewhere. Anywhere from 0.05 to 0.20 sec is not unusual to lose at each >move when playing on a server, even with timestamp. Yes. I also have seen it with Winboard. However, one has to be careful, to measure correctly. ICS games are typically played with ponder on. When you use the polling method for detecting input, you will lose some fractions of a second there (of course the amount depends on the frequency, you poll for input). You may start your internal reference clock not at the moment, you receive the move, but after you have parsed the move. Perhaps you undo a wrong ponder move first, too. Perhaps more. Similar, you may stop the clock after you found the move. Before you send it, you might format and output some PV, ... With winboard, one should turn the move animation off, for fast games. For ICC games, one should leave more than a second of safety margin, because ICC can send wrongly rounded time stamps. When scaling things linearily, in fast games (say 1 1) an engine might use less safety margin. It could believe, that even when things go bad, never more than 1/2 or so will be lost. But it happens, nevertheless. Regards, Dieter PS. My engines uses a too small safety margin ... 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.