Author: Anthony Cozzie
Date: 22:55:00 03/01/03
Go up one level in this thread
On March 01, 2003 at 16:39:50, Mike Hood wrote: >On March 01, 2003 at 14:51:15, Mike S. wrote: > >>On March 01, 2003 at 14:14:51, Mike Hood wrote: >> >>>The free Chessbase engine List 5.04 seems to be a very strong engine, probably >>>the strongest engine that doesn't use endgame tablebases. I wonder how much >>>stronger it would be if the programmer added tablebase probing...? >> >>I guess +10 Elo (max.). >> >>Whenever I read results of people who took a close look at the end of many >>computer games (which wasn't reported often), to see if and how tbs. were >>responsible for the results, the baseline was that these are very seldom cases. >>For example, this was examined when SSDF tests were affected by the incomplete >>problem, and it turned out IIRC that although it was there, it obviously didn't >>affect results at all (or at least only very few; I don't remember exactly). The >>problem was solved anyway. >> >>Nevertheless, tablebases are an improvement of course, because the engines will >>most often play more precisely in the end, disregarding if the result has >>already been decided earlier in the game. >> >>Note also that when using Lust 5.04 (or any other engine) in a Fritz GUI, the >>"static" tablebase access will always be done by the GUI as soon as a tbs. >>positions in reached in the game. >> >>Currently I'm not sure which one is stronger, List 5.04, Ruffian 1.0.1 or SOS.3 >>f.Arena. For example, see this rating list compiled by K.Wlotzka (he runs games >>at 10m+10s, A12000, with 10 predefined opening variants): >> >>http://wlotzka.bei.t-online.de/aktuelle_rangliste.htm >> >>Regads, >>M.Scheidl > >I'm surprised that List is higher than Ruffian. But I have to admit, I only have >one PC, so my engine-engine matches without pondering don't give a definitive >answer, just an estimate. > >The use of EGTBs ought to give engines a speed boost as they approach endgame >positions. (If the EGTB data is used correctly, which isn't always the case, as >I've previously shown in http://www.talkchess.com/forums/1/message.html?283960). >When during a search a position occurs that is in an EGTB the value can be read >and the search along that branch terminated, whereas if the EGTB is not present >the engine has to spend time searching further along that branch. this is true, but remember that a disk access will take ~10ms on a 7200RPM disk. Fritz searches about 1M nps, so in 10ms it can search 10,000 normal nodes. . . Moreover, most of the time the game is already won by the time it reaches the tbs. I will probably build in support for tbs, but most of the time I plan to use interior node recognizers in Zappa. anthony
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.