Author: Mike S.
Date: 17:07:02 05/07/03
Go up one level in this thread
On May 07, 2003 at 17:06:59, Robin Smith wrote: >On May 06, 2003 at 15:36:01, Mike S. wrote: > >>This is an interesting way to check opening novelties btw. :o)) >Are you saying position learning makes automated checking for opening novelties >more effective? Can you share more details? The idea is simple: Due to the position learning function, engines which are capable of that, will try to improve their game (avoid moves which turn out to be bad), game by game. Of course the method will be most useful to test sharp tactical opening ideas. (For a positional type of position, other analysis methods may still be sufficient or even better, i.e. "correspondence analysis" function with long calculation times and predefined depths and number of variants and sub-variants etc.) I've run long matches with 24+ (up to 60) games each, from the same opening variant always (using an "opening database" containing only that single variant). Unfortunately that wasn't a sharp or trap-like variant, so I don't have convincing examples available from that. But some years ago, I was able to repeat the following - incorrect! - opening trap against Genius 4 unlimited times: [Event "40/20:00 P133"] [Site "Wien"] [Date "1998.??.??"] [Round "?"] [White "Scheidl, M."] [Black "Genius 4"] [Result "1-0"] [ECO "B02"] [Annotator "Scheidl,M"] [PlyCount "25"] [EventDate "1998.??.??"] 1. e4 Nf6 2. e5 Nd5 3. Nf3 d6 4. Bc4 Nb6 5. Bxf7+ Kxf7 6. Ng5+ Kg6 7. Qf3 Kxg5 8. Qf7 g6 9. d3+ Kh5 10. Qf4 h6 11. g3 Bg4 12. h3 Qd7 13. hxg4# 1-0 The wrong move is 7...Kxg5. Later, I could beat Genius 5 with this too - but only once... :o) Of course today, most engines will see soon that 7...Kxg5 is a blunder. But it illustrates the principle. Regards, M.Scheidl
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.