Author: Sandro Necchi
Date: 23:17:41 11/27/01
Go up one level in this thread
On November 27, 2001 at 08:10:02, Peter Berger wrote: >On November 27, 2001 at 07:15:14, Sandro Necchi wrote: > >>Maybe the engine is only 40-50 points stronger than 5.32, but with a better book >>(this time made by myself again. 5.32 book is not made by me) and with a better >>learning feauture seems to be at least stronger than previous 5.32 by 80 points. >>This is what came out from test games. More than 600 made by me and many other >>made elsewhere. >>This is what I know. You can believe it or not, it only depends on you. >> >>Ciao >>Sandro >> > >Hi Sandro, > >this is a topic of interest to me. > >So you say we have: > >better engine: 40-50 points >better learning feature : ? - OK, let's say 10 points ? > >So your special book over a default GM book ( 5.32) is 20-30 points .. ? > >IMHO that's good work, others probably disagree. To be honest I do not care if people believe me or not. Who knows me knows that I am sincere and I say what I know even if I can make mistakes too. Since last year I did not test the commercial book as I wanted to spend more time preparing for the next tournament, this year I did the opposit spending 3 full months on the commercial book to make it good (fun) as I wanted. Well, to say things is easy, to make them non so:-)) This is the best I could do for the time being. The main target was to make a program (engine + book) which is really fun to play for humans, not to score better against other chess programs. However, making it better it does score better against other chess programs. Over 100 points was the last picture against other chess programs (120 to be precise; last score Shredder 6 - Deep Fritz 1h games - 38.5 - 23.5 +32 =13 -17), but I normally think it is better to use lower figures, so I say probably 80 points in total. > >But another question: how about your book against no book at all ? No, I am a openings book fanatic, so to me to test a program without book is REALLY no sense. >( Few tests >on this topic : recently someone did Yace against Yace with no book and the >no-book version even won the contest). I do not think this would happen with Shredder 6. > >Did you do any tests in this field or were your above numbers guesswork ? No, I study the engine and after I make the book for it. The idea is to avoid positions where the program does not know what to do or play weaker and let it play where it does play better. So, it is extremely easy... Since I start working on a book which is not the ideal in many variations, this gives a picture which is probably correct to estimate the engine improvement alone. > >Regards, >pete Regards Sandro
This page took 0.02 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.