Author: Anthony Cozzie
Date: 12:05:12 08/25/03
Go up one level in this thread
On August 25, 2003 at 13:51:10, andy ********* wrote: >On August 25, 2003 at 07:55:52, Mark Young wrote: > >>On August 25, 2003 at 07:13:54, David Bigler wrote: >> >>>Hi all !! >>> >>>My computer often plays on PlayChess.com >>>My hardware system is 1 Athlon XP 2600+ with 512 Mb RAM >>>I use 48 mb hashtables and no tablebases because the programm I use is Shredder >>>7.04 and therefore, tablebases slows too much this programm in blitz. >>> >>>Shredder 7.04 is realy an impressive programm. And it is far the strongest one >>>for the moment. >>>I actually have an ELO of 2860 and my record was 2985 on a single CPU. >>> >>>The secret of this high ELO with a single CPU is that I work a lot on the >>>opening book. >>>I make the computer play my one tournament opening because it is more >>>interesting also for me. Then each game the computer plays, I analyse it and >>>modify the opening book if necessary. >>> >>>It seems that now even computers with ELO of arround 2500 ELO often uses 2 CPU. >>>So the level is going higher and higher. >>> >> >> >>>My target is to have over 3000 ELO with this single CPU. I am not sure it is >>>possible...........but we will see. >> >>A 3000 blitz rating is meaningless on playchess.com. Anyone with the same or >>close to the same hardware and program you are running is playing the same >>strengh you are playing no matter what rating they show on playchess.com. You >>can be 3000 rating and they could be 2600 rating. Still you are not really 400 >>rating points better, just using a better formula most likely. > >Yes with identical hardware and settings , both programs must have the same >intrinsic playing strength , however with a more optimised book it would be >possible to have an higher rating. The problem with very high ratings on >playchess.com is often the player's "incestous" formula , often playing the >same people time after time, perhaps creating a kind of isolated grading pool ? >I experimented with 2 accounts , which had the same hardware,settings and book. >The "no rating restrictions" had a grade of approx 2500 , while the more > "restrictive" was 2700+ , that's clearly nonsense ! Part of this is also the fact that the elo system itself is not perfect. Sonas has done some stats on this at some point I think (his work with humans is a lot better than his work with computers) anthony
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.