Author: Dann Corbit
Date: 19:51:49 04/14/05
Go up one level in this thread
On April 14, 2005 at 14:25:41, Ted Summers wrote: >On April 14, 2005 at 10:50:14, Uri Blass wrote: > >>On April 14, 2005 at 08:42:37, Ted Summers wrote: >> >>>Is there a website that shows the Top Performers of Non Tablebase Tournaments. >>>Would like to see which engines gain the most from the use of tablebases. Since >>>I know all engine gain from the use of tablebase, but do some gain more then >>>others? >> >>You know wrong. >>There are engines who lose from tablebases because of bugs and there are many >>engines that do not use them. >> >>Uri > >I disagree with your opinion. Overall engines I beleive gain from the use of >Tablebases. Every experiment I have ever seen shows the contrary: Tablebase files are neutral. There is neither a gain nor loss in Elo. Bitbase files (on the other hand) do lead to an increase in strength. For example: ftp://cap.connx.com/pub/crafty-tablebase-experiment.doc http://chessprogramming.org/cccsearch/ccc.php?art_id=307405 http://f11.parsimony.net/forum16635/messages/51417.htm >I think that is why alot of engine writers include support for the >use of tablesbases. I think that they hope for greater strength, but are more likely to do it for art. A 45 move checkmate with every move perfect is pretty to look at. >Lastly of course there are engines that do not use them, >this I already knew. I have never seen any experiment show a strength increase from the use of tablebase files. Neither have I ever seen any loss of strength demonstrated. It appears that the cost of a lookup is about the same as the cost of calculation. A tablebase search only provides an optimal path. Most good chess engines will be smart enough to find the checkmate anyway. The odds that a chess engine will run around and around the mulberry bush while never finding the mate appear to be so low as to be immesurable. On the other hand, all the tests I have seen involved strong engines. Maybe they do add significant strength to weak engines. On the other, other hand -- I expect that people who write weak engines will also do the probes in a bad way and probably cost themselves even more from that.
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.