Author: Mike S.
Date: 14:46:35 04/07/02
Go up one level in this thread
On April 07, 2002 at 14:41:20, Mike Hood wrote: >(...) >This is a matter of opinion. I think that today's chess programmers have the >right to assume the use of 5-piece tablebases. I would strongly recommend *not* to rely on tablebases, at least not for major parts of endgame knowledge (i.e. bacic mates). The assumption that everybody uses 5-piece tbs., is wrong: First, not many people, except computer chess fans, even know what tablebases are. Tablebases are expert's stuff. Second, 5-piece tablebase are not always an advantage. I even doubt that they are an advantage in more than, say 60% of practical positions, when accessed during the search. As a consequence of that, I use 4-piece tablebases only +RB-R, RN-R. (For 5-piece endgame analysis, I still have the Thompson tables available.) A possible disadvantage may sometimes not be visible, because both sides are affected. Third, AFAIK most programs don't access tbs. in the quiescence search (AFAIK only Nimzo does that with it's NCD tables). Which means, missing knowledge can not be replaced in the quiescence search by using tbs. I think it is a big danger for (professional) programmers, to adapt too much to the type of users who are computer chess experts. This is an audience of a limited number IMO. If your program is sold in big shops and can't do something without 5-piece tablebases, 10 of 100 customers will say the program can't do it at all (5 will use tbs., and 85 will not notice it at all :o). But OTOH I think, most major programs have much built-in endgame knowledge, and do not rely on tablebases too much anyway. 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.