Author: Mike Hood
Date: 05:09:02 06/09/03
Go up one level in this thread
On June 08, 2003 at 20:41:42, Dave Kuntzsch wrote: >Just got a new computer that is large enough to hold the five piece table bases. >Do they appreciably improve ELO if the program is already relatively strong? Or >are they only good for tricky endings that occur infrequently? Has anyone done a >same engine study with one using the four piece, and the other the five piece? >I'm currenty downloading them, just in case. > >Dave I would answer your question with a definite Yes. The importance of endgame tablebases for games where an EGTB position is on the board is a lesser factor. It might be good for practice to play King/Queen/Pawn vs King/Queen, but (1) in championship games positions with 5 pieces rarely arise and (2) apart from a few glaring examples, strong players -- including strong chess programs -- can play these endgames without EGTB knowledge. The main advantage of EGTBs is in positions with slightly more than 5 pieces on the board, let's say 8 to 10 pieces. Positions with 10 pieces arise in close to 100% of engine-engine games between opponents of similar strength. The advantage to a chess program that uses EGTBs is immense. Instead of "only" having to search for Mate or increasing his material advantage, an engine can end a branch of his search tree if he encounters a line that by exchanging (or sacrificing!) material leads to a position that according to the EGTBs is a forced mate. If this is just a side branch, it means that this branch can be discarded and the processor time used more productively for looking at other branches. If it is the main branch, the engine can try to force the opponent into this EGTB position. To quantify the improvement... I remember someone posting the results of a Crafty vs Crafty tournament in here, with and without tablebases, and he claimed there was a 25 Elo point improvement. That might not sound like much, but I'd be glad if I could improve my own playing skill by 25 points! All of the top chess programs use EGTB probing. That's an argument in itself, isn't it? If a chess programmer thought EGTBs wouldn't improve his engine's performance he wouldn't add it. The strongest engine that I know that doesn't use EGTB probing is List. I'm curious whether the next version will add EGTB probing, or whether the author will make a valiant effort to prove to the world that he can be one of the best without using EGTBs. Let's wait and see. (Or if Fritz Reul is reading this, maybe he can comment). I'm sure that people can produce examples of positions where an engine with EGTB probing can find a win that an engine without probing can't find. There must be even more examples of positions where an engine with probing can find a draw in a seemingly lost position that an engine without probing can't find. Apart from obscure positions where castling is still valid in endgames, I've never seen an example where an engine without probing makes a better move than an engine with probing.
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.