Author: Gian-Carlo Pascutto
Date: 04:59:12 04/23/02
Go up one level in this thread
On April 23, 2002 at 04:43:00, David Dory wrote: >Yes, I still believe Hiarcs7.32 got stronger relative to other programs when >playing at time controls longer than blitz. I've played other engines that were >that way, also. Dann C. and another poster mentioned their experience with Amy, >which certainly did get stronger with longer time controls, relative to other >programs. I have already said in previous posts that I do not think Amy is a fair example, because the program is just broken in fast timecontrols. Dann played a 100 games match Sjeng-Amy at 1 0 and Sjeng won 99.5-0.5. This is not a normal result nor is it a result that is caused by one engine being relatively better/worse as timecontrols get faster/slower. It is a result of a buggy vs a bugfree engine at the tested timecontrol. There are some engine that have a fixed delay for each move. These will also be relatively worse than engine that don't have it in fast games. I discount such effects when talking about these issues. i.e. I agree there will be differences in performance when one of has effects, but when they don't, I don't think engines will play significantly (relatively) stronger or worse with changing timecontrols. Most professional/well tested engines will have neither of those problems. >>The fact that Palm Tiger, despite an > 100x NPS advantage, can still compete >>with some strong amateurs at blitz. > >I find it impossible to believe that Palm Tiger has a > 100x nps advantage over >most good PC based programs. Don't you mean Palm Tiger has > 100x nps >DISADVANTAGE? Yes, of course :) >Anecdotal evidence is indicative, but not something you can rely on without the >imposition of rather strict standards of testing. This isn't always possible, >but I'm sure you would be the first person to say "Wait a minute, this is just >heresay evidence", in re-reading your paragraph regarding the amateurs and the >Palm Tiger. I'll keep it in mind, however. I agree. I would like reliable data to be able to determine scientifically whether this difference with timecontrols exists or not. I would very much like to know it, if I am wrong, because this will influence how I test my engine. But such data is not availble, so I use the best I have, which indicates no such effect exists. >I'm unsure why the idea that a faster chess engine would be slightly better at >a blitz time control is not somewhat intuitive. I agree this is intuivite, but in computer chess intuitive is often wrong. Do not forget these 'speed differences' are caused by something, and this something will have an effect in both blitz and slow games. Don't underestimate knowledge. -- GCP
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.