Author: Terry McCracken
Date: 01:25:04 07/29/03
Go up one level in this thread
On July 29, 2003 at 03:25:02, margolies,marc wrote: >terry, >the example you have given in my opinion is not a table base related issue. >If the egtbs are complete, it is my understanding-specifically with shredder >from its helpfile doc- that in calculating a line of variations (forcing say) >that leads to a table base position, then the engine will lookup that evaluation >--in your example a draw-- and assign it to the line. >The kind of problem you describe is the result of machines trying to calculate a >solution that is already known in real time, not a tablebase problem. >When running a chess camp with my coach 6 years ago I remeber Justin Sarkin, now >an IM getting into a position in Hiarcs on my Mac laptop at the camp where he >was losing to HIARCS in a technical King and Pawn ending. We all knew this >ending because it was the subject of a lecture that day by our EG specialist, >Leon Verhovsky who wrote the book "Nichiya!" But HIARCS could not convert the >extra half point and the game was drawn. >In short, I do not think your example butresses your opinions. > >Admittedly, it is natural to resist putting 8 gigabytes of endgame tables on an >older machine. I know that win95 probably cannot address 8 gigabytes of rom for >example. But this is the newest programs need. It is also relevant to use faster >access on these drives. To defrag them regulrly etc. These are performance >indicators which players do not always elaborate upon. > >Uri, tha > > >On July 29, 2003 at 03:05:57, Uri Blass wrote: > >>On July 28, 2003 at 23:09:42, Terry McCracken wrote: >> >>>On July 28, 2003 at 20:25:51, margolies,marc wrote: >>> >>>>Ok Terry, what are the problems that you've seen caused by implementing EGTBs >>>>and removing knowledge files from engines? I think the conversation is heading >>>>somewhere now... >>>> >>>I've seen programmes miss for example, combinations that lead to a forced draw >>>with a pawn or pawns on the rook file with the wrong coloured Bishop. >> >>Note that the original subject was about KBN vs K and I see no demage in playing >>strength if you know to win it only by tablebases. >> >>I did not implement it in movei and the practical importance of it is small and >>one of the reasons is that computer opponents with tablebases usually prefer to >>resign or to let movei to get a bigger material advantage. >> >>Uri As I said guys I'm bias in this area of computer chess....and although I think EGTBS are a great idea, and practical, I'd like the computer programme itself know this....even the old Mach III Master had this type of knowledge and it was a slow computer...16Mhz, 16 bit processor (Motorola 68000) with only 8 bit code!:) Also in those days, the programmes were in machine language, assembler, which only Ed Schroder I believe still employs at the heart of Rebel....I still think this method is superior, but painstaking. Also I believe Christophe worked hard on endgame knowledge as he didn't employ EGTBS until CT 15 which has both a good endgame and now includes tablebases to top it off:) Yes I did notice EGTBS not come into play fast enough when a middlegame position had to be calculated and involved a series of sacs to reduce it to a position EGTBS would come into play....it was rather disappointing to see fast hardware slow to find the draw....when slow hardware with an outdated programme was very fast! It's "Old School", and I guess I'm getting old as well....sigh...not their yet mind you!;-) Nevertheless, a little extra knowledge goes a long ways and may be worth a few plies! Terry P.S. Ed and Christophe are the experts...maybe they can voice their opinion on this matter?
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.