Author: Dann Corbit
Date: 08:48:48 01/07/02
Go up one level in this thread
On January 07, 2002 at 09:40:17, K. Burcham wrote:
>we see elo opinions here often.
>but how does an elo difference convert to win/loss.
>
>using the same program. same default settings.
>one program has 1000mhz 256 megs ram.
>second program has 2000mhz 512 megs ram.
>estimated kns for program one, 500kns.
>estimated kns for program two, 1000kns.
>
>if we play a 200 game match.
>this is time setting per move. percentage of wins for 2000mhz.
>
>1sec 3sec 6sec 10sec 15 sec 20sec 30 sec 45sec 1min 2min 3min 5min
>
>95% 93% 91% 89% 87% 85% 83% 81% 79% 77% 75% 73%
>
>anyone think this chart would be close
Here are some figures at slow time control (equivalent to G/4 hrs for the
equivalent hardware used below), for engines where I have run quite a few games:
Program Elo Games diff
Yace 0.99.50 2623 176 0
Amy-07 2546 133 77
ZChess-222 2515 214 108
Amyan-1423 2375 112 248
Fortress-162 2283 165 340
Faile-14 2231 186 392
31337-30 2054 180 569
Here is a 5 minute blitz rating list for the same engines:
Program Elo Games diff
Yace 0.99.50 2524 456 0
ZChess 2.22 2428 520 96
Amy 0.7 2329 768 195 {fits my theory that Amy gains w/ time}
Amyan 1.423i 2328 340 196
Fortress 1.6.2 2163 768 361
Faile 1.4 2135 768 389
31337 0.30-2 1932 520 592
Now, the absolute numbers are irrelevant, but the differences should be telling.
At 48x longer time control the differences in ELO are not dramatic.
In other words, I would expect the ratios to hold approximately the same (with a
bit of variation here or there) for all time differences.
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.