Author: Sandro Necchi
Date: 23:16:14 05/31/01
Go up one level in this thread
On May 30, 2001 at 13:57:54, Dana Turnmire wrote: > This article appeared in the 1996 issue of Computer Reports and was written >by Michael Byrne. It shows me the only fair way to find an engines true >strength is WITHOUT the opening book. It also shows why Deep Blue did so well >against Kasparov. When you have a killer book written by a grandmaster >(DeFirmian) especially suited for play against Kasparov and the fact that >Kasparov was playing blind and not even allowed to look at any of Deep Blue's >games the results were tainted. > The same goes for any commercial program playing against grandmasters. Take >the program out of the box and use the default settings and book to find out if >it can really beat a GM. I always believed that a very strong opening book should have been essential to allow a chess program to reach a very high strenght. This is why I started to make openings book for chess programs in 1978. > > M-Chess Pro 5.0 "Hidden Book" (by Michael Byrne) > > "I recently played this game between M-Chess Pro-90 mhz Pentium and Mephisto >68030 33mhz dedicated unit. The time control was 3 hours for the game. What >made this game interesting was the discovery of a 'hidden' book in M-Chess >Pro5.0. The settings were full strength with tournament book ON. M Chess Pro >5.0 stayed in book for 39 MOVES. To me that was unbelievable - and also it is >worth noting that on 17.Kb1 - Genius 68030 was out of book. The game is >essentially over at move 39 - any top program could mop that mess up left on the >board. M-Chess Pro used 1 second for 39 moves! Obviously Marty did a good job >of 'tuning' his book against the Genius book. I (to improve the book was my job and not Marty's) did not make a specific work to beat Genius, but I included in the book all good moves which were discorvered, like a GM would include in it's repertoir all its findings. It's the same phylosophy. I did not want to cheat. I hope you understand this. >The 68030 book by the way is >almost exactly the same as the Genius 2 tournament book. Richard Lang could >confirm if it is the same. > I wanted to see how M-Chess Pro would do on its own after 17...a4. > I replayed the position up to that point, turned the M-Chess book off - and it >continues to make the same moves instantaneously right up to the move 39. > You can't turn this book off - I even renamed the book p these moves are in >the program! There is a feature in the program that allows the program to create an add on book which automatically stores all the good played moves, so once it saw them they are added to its memory. I think this is a great feature which makes the program closer to a human. > [Event "G/3 HRS."] > [Site "Philadelphia, PA"] > [Date "1995.12.12"] > [Round "1"] > [White "M-Chess 5"] > [Black "Genius 68030"] > [Result "1-0"] > 1.e4 c5 2.Nf3 e6 3.d4 cxd4 4.Nxd4 a6 5.Bd3 Bc5 6.Nb3 Ba7 7.Qe2 Nc6 8.Be3 >Bxe3 9.Qxe3 d6 10.Nc3 Nf6 11.0-0 0-0 12.f4 Qc7 13.Rhg1 b5 14.g4 b4 15.g5 Nd7 >16.Ne2 a5 17.Kb1 a4 18.Nbc1 a3 19.b3 Bb7 20.h4 Ne7 21.h5 Qc5 22.Qh3 Rfe8 23.g6 >hxg6 24.hxg6 Nxg6 25.Rg5 Qf2 26.e5 Ndf8 27.exd6 Nxf4 28.Nxf4 Qxf4 29.Rdg1 Qd4 >30.Rxg7+ Qxg7 31.Rxg7+ Kxg7 32.Qg4+ Kh8 33.Qxb4 Bd5 34.Qh4+ Kg8 35.Qg5+ Kh8 >36.c4 Bc6 37.Qf6+ Kg8 38.Ne2 Ra5 39.Nd4 Bb7 40.Nb5 Bc6 41.Qg5+ Kh8 42.d7 Bxd7 43 >Qf6+ Kg8 44.Nd6 Rb8 45.Qxf7+ Kh8 46.Qf4 Kg7 47.Qg3+ Kf6 48.Ne4+ Kf7 49.Qxb8 Ra6 >50.Qf4+ Kg7 51.Qf6+ Kg8 52.c5 Bb5 53.Bxb5 Ra7 54.Qg5+ Rg7 55.f6+ Kh8 56.Qh5+ Rh7 >57.Nxh7 Nxh7 58.Bd3 1-0 > > Now, I am really curious - how well is MCP5 doing against the Genius programs? > MChess Pro5.0 Pentium 90 MHz, 2458 > > Genius 3 P90 13-7 Rebel 6.0 P90 16-4 > Hiarcs 3 P90 11.5-1.5 MCPro 4.0 P90 5-7 > Rebel 7 486/66 9-8 Genius 3 486/66 11.5-8.5 > Genius 2 486/66 15-5 WChess P90 6-14 > MCPro 40 486/66 12-8 Rebel 6 486/66 16-4 > Genius 68030 6.5-2.5 WChess 486/66 14.5-5.5 > CM30 King2.0 4-2 ChGenius 486/66 22-8 > MCPr35 486/66 3-1 Fritz 3 486/66 1-1 > Kallis 486/66 16.5-3.5 SPARC 20 MHz 4.5-2.5 > Meph. RISC 1-0 Chess M. King 3-0 > > Look at that - - 21.5 - 7.5 against Genius 2 (w/Genius 68030) and 13-7 against >Genius 3. > And what is more amazing - it's only 5-7 (under 50%) against MCP4! MCP4 is >only 6.5-13.5 against Genius 3. > What's my point -- it appears that Marty has identified significant weaknesses >in the Genius programs' opening books - while not greatly improving the actual >playing strength of his program - note also the paltry 12-8 record against MCP4 >on a 486/66 machine! I do not agree. MCP 5.0 was weaker tactically, but improved positionally and in the endgame, so it may suffer the better tactics of MCP4 in some games/variations. Sandro Necchi >Granted, the number of games here are not statistically >significant - but this may be of interest to consumers/researchers and to those >who want to discover gambusting lines against Genius!"
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.