Author: Uri Blass
Date: 21:50:45 04/25/05
Go up one level in this thread
On April 25, 2005 at 20:50:59, Dann Corbit wrote: >On April 25, 2005 at 20:34:19, pavel wrote: > >>First 1000 Games: >>-------------------------------------------------------------------------------- >> Program Elo + - Games Score Av.Op. Draws >> >> 1 Aristarch 4.50 : 2519 19 19 1000 55.4 % 2481 23.3 % >> 2 Yace Paderborn : 2481 19 19 1000 44.6 % 2519 23.3 % >> >> >>Individual statistics: >> >>1 Aristarch 4.50 : 2519 1000 (+437,=233,-330), 55.4 % >> >>Yace Paderborn : 1000 (+437,=233,-330), 55.4 % >> >>2 Yace Paderborn : 2481 1000 (+330,=233,-437), 44.6 % >> >>Aristarch 4.50 : 1000 (+330,=233,-437), 44.6 % >>-------------------------------------------------------------------------------- >> >>All games (so far 5004 games): >>-------------------------------------------------------------------------------- >> Program Elo + - Games Score Av.Op. Draws >> >> 1 Aristarch 4.50 : 2516 8 8 5004 54.7 % 2484 23.9 % >> 2 Yace Paderborn : 2484 8 8 5004 45.3 % 2516 23.9 % >> >> >>Individual statistics: >> >>1 Aristarch 4.50 : 2516 5004 (+2140,=1196,-1668), 54.7 % >> >>Yace Paderborn : 5004 (+2140,=1196,-1668), 54.7 % >> >>2 Yace Paderborn : 2484 5004 (+1668,=1196,-2140), 45.3 % >> >>Aristarch 4.50 : 5004 (+1668,=1196,-2140), 45.3 % >> >>-------------------------------------------------------------------------------- >> >>At one point I thought the positional learning was really helping YACE (it did >>good in 3rd and 4th 1000 games). But now it seems that the overall differance is >>negligable since the first 1000 games. I will run another 5000 than I will stop. >> >>pavs > >It is interesting that you chose two programs that are very close in strength. >At this point, it is still hard to be sure which one is actually stonger. So in >this particular case, the entire experiment will be needed to resolve the >question. I can be sure that Aristarch is clearly stronger based on other data like CEGT see http://www.husvankempen.de/nunn/cegt.html but even based on the data that is posted it seems that Aristarch is clearly stronger. 1 Aristarch 4.50 : 2516 8 8 5004 54.7 % 2484 23.9 % 2 Yace Paderborn : 2484 8 8 5004 45.3 % 2516 23.9 % The difference is 32 elo and the error margin of the difference is clearly smaller. Note that I think that a match of thousands of games between 2 programs is a very bad way to test which program is stronger because you may find by that way that A can score 54% against B,B can score 54% against C and C can score 54% against A. Uri
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.