Author: Mike S.
Date: 17:31:11 12/05/02
Go up one level in this thread
On December 05, 2002 at 11:28:53, Louis Fagliano wrote: >(...) >I wonder how many games (out the the hundred thousand or so) that the SSDF has >played ended like this. This is obviously one of the problematic details of SSDF testing which were critized then and when (by me too, among others), but without any chance that the'll ever change their practise. If this is a case, and I guess it is, caused by using the Fritz Endgame Turbo (an incomplete selection), than that's it. It's an existing product, they have it, they use it... no matter if most expert user's just don't do it like that. It seems to me that KQB vs. KB was missing in that game, while KBP vs. KB was availabe, so Black didn't want to "leave" the tbs.; a typical incomplete problem. I analysed with Shredder 5.32 and 4-piece tbs. only, and here it always wanted to promte the pawn from move 86 on with no problems visible. SSDF have sometimes pointed to their policy of strictly using programmer's recommendations (or manufacturers recommendations), but IMO that's a decision which is neiter obvious for an *independent* testing group, nor does it take into account that such recommendations can be compromises (with customers who are not as good equipped as the average computer chess fan and who may not have advanced user's knowledge). They offer a list to the computer chess public, but often don't test in a way the majority of the computer chess public uses the programs (nor the programmers; I don't assume programmers use incomplete tbs...). I don't think it's a big problem though and won't have significant influence on the results... but it's an unnecessary fault so to speak. That it's very good when a program handles incomplete sets correctly, is another topic... I think Crafty is capable of that. But normally, either complete sets of the 3- and 4-piece tbs., or the 3/4/5-piece tbs. will be in use, or at least using tbs. with pawns without having all those which can arise from the pawn (under-)promotions, will be avoided. For example, I use 4-piece + RB-R, RN-R: no incomplete problem. We must remember to be grateful, that the SSDF people do all the work to compile the list, despite a number of unsatisfactory details to keep in mind. After all, nobody is perfect. I guess nobody could convince SSDF using i.e. only *complete* 4- or 5-piece tbs. Regards, M.Scheidl
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.