Author: Jonas Cohonas
Date: 05:07:13 11/07/05
Go up one level in this thread
On November 07, 2005 at 02:01:03, Eduard Nemeth wrote: >On November 07, 2005 at 01:53:16, Kurt Utzinger wrote: > >> Hi Eduard >> After more than 2000 games, Toga II 1.0 >> seems 42 Elo weaker than Fruit 2.2.1 >> http://www.husvankempen.de/nunn/rangliste.html >> Why do you have any doubts about this fact? >> You can't conclude something on the basis of >> some (dumb) blitz games on Playchess.com >> Regards >> Kurt > >1. For me gives LIVE games with tuned books more informations about the Strength >of both engines, clear more than matches with only 5 move-book maches or another >short openings like Nunn-Test etc.. > >2. on playchess will played with permanent brain on very fast computers! That is >for me an very important point! > >Best, >Eduard. I agree with Kurt and your argument about permanent brain + fast hardware on playchess is just, well silly... for example, at 3 0, lets just say that for arguments sake that the computers at playchess are 3 times faster than your average PC for testing in the chess community and that all those uses ponder=off then a 3 0 game would be (at best) roughly the same as 18 0 on a regular PC (ponder=off), which is still very short time controls. You say that the fact that the games are LIVE plays a part in the information you get about the engines strength... Do you understand that the shorter the book lines, the more information you get about an engines actual playing strength? Do you realize that with these "tuned books" you mention, you are at risk of observing engines of equal strenght where one keeps losing because one has a much better book or one has a very bad book? Not to mention a lot of times you will get a lopsided result from engines of equal strength and equal strength books because one has better hardware? Then there is lag issues, misleading results based on poorly setups in terms of hardware and software, how do you know that the engine losing is not the cause of someones virus program stealing all the processor time and so on. Playchess in my opinion gives very unreliable results (when it comes to eng-eng matches) as it is very far from being a controlled inviornment as opposed to the excellent work of testers like Kurt, Graham, CEGT team, SSDF etc. The sheer bulk of games on playchess will give a good estimate, but to base anything about an engines strength from a few blitz/bullet games from playchess is just nonsense if you ask me. Regards Jonas
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.