Author: Robert Hyatt
Date: 07:44:31 08/20/05
Go up one level in this thread
On August 20, 2005 at 09:54:11, Tord Romstad wrote: >On August 20, 2005 at 09:20:43, Robert Hyatt wrote: >>I do have a problem with programs that are using shared code for tasks >>that are not strictly one-answer tasks. EGTBs are a counter-point, in that >>everyone gets the _same_ EGTB score whether they use Eugene's code, write their >>own probe for his tables, or write code to build their own tables. > >It is not quite that simple. First of all, it is not at all sure that everybody >gets >the same EGTB score regardless of what code and tables they use. The score >will depend on whether you use distance to mate or distance to conversion, >on whether you consider the 50 move rule, and on whether you have bugs >(we are talking about non-trivial code, after all). The point is that this is a one to one mapping function. One position maps to one endgame table entry. Whether that entry is a mate in N, or a conversion in N is irrelevant. Just as surely as we should all be free to use things like abs() and sqrt() which are one to one mappings. But if it is _ever_ a one to many mapping, then the mapping algorithm itself is influencing the result, such as choosing between move A and move B. There is much that could be used to make that decision. Over results of each move in the games previously played. Ratings of the two players. Age of the games. Human suggestions for which is better. learning results. etc... > An even more important >point is that EBGT probing in not only about getting the right result, but also >about getting the right result *quickly*. It is not at all easy to compress the >EGTBs down to a manageable size and still be able to decompress and probe >quickly on the fly. Again, this is not chess-related. Compressed or uncompressed does not matter one iota to the chess engine, nor to the move it will make. > >In my opinion, including Nalimov's EGTB code in the program is no better >(and no worse) than using a GUI book. > >Tord We have something to disagree about then. :)
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.