Author: Uri Blass
Date: 17:16:06 10/22/03
Go up one level in this thread
On October 22, 2003 at 19:33:03, Mike S. wrote: >On October 22, 2003 at 09:52:29, Uri Blass wrote: > >>(...) >>If aristarch has better branching factor than Fritz5.32 then it proves nothing >>because Fritz5.32 is an outdated engine. > >Coincidentially, I've included both engines in a test for the branching factor, >IOW factor = total time after ply x / total time after ply x-1. (Im not sure if >we have the same definition.) > >ply depth: 10 11 12 13 14 15 16 Ø >-------------------------------------------------------------------------------- >Fritz 7 2,9 2,1 2,4 3,7 2,8 >Fritz 5.32 3,0 4,9 2,7 >> 3,5 >Chess Tiger 14.0 3,1 3,0 2,2 2,3 2,7 >Hiarcs 7.32 3,9 5,4 2,4 3,9 >Comet B48 3,4 1,8 2,6 2,6 >Shredder 5 4,8 2,9 2,1 2,5 3,1 >Yace 0.99.56 1,9 3,7 2,7 2,8 >Aristarch 4.4 2,3 4,4 >> 3,4 >Crafty 18.15 2,7 2,9 2,0 2,5! >-------------------------------------------------------------------------------- > average 3,0 > >(tested on Athlon@1,2 GHz, 256 MB Hash, from the starting position*) > >*) later I realised: To use the starting position wasn't the best choice, as the >number of possible moves in the first plies is not typical... > >But the factor was very similar for Fritz 5.32 and Aristarch 4.4 (which isn't >the latest version though)... Unfortunately, there are not enough values for a >reliable result. It was only a quick analysis done in short time. > >I have no opinion or results, how this factor is related to the strenght in >blitz vs. long time controls... for example, I'm not aware that Hiarcs 7.32 >(3,9) was particularly worse in long time controls compared to blitz, nor that >it was vice versa with Crafty 18.15 (2,5!). Although it would seem logic that an >engine with a smaller branching factor can gain more depth from additional time. There is a problem to compare different engines in this way because of different pruning and extensions. What I meant by better branching factor is not the exact definition and I simply copied it from christhophe who said the same about Genius. I know that comparing the branching factor of programs without the same extensions and the same pruning may be misleading and the only real test is games. I can add that hiarcs7.32 is known to be worse in long time control. I remember from analyzing correspondence games that both fritz5 and Hiarcs7.32 had a very bad branching factor with long analysis and I guess it is simply some bug(for example having some overflow that cause wrong order of moves). 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.