Author: Sandro Necchi
Date: 10:38:11 02/21/05
Go up one level in this thread
On February 21, 2005 at 04:25:13, Vasik Rajlich wrote: >On February 20, 2005 at 13:20:44, Sandro Necchi wrote: > >>On February 20, 2005 at 13:02:51, Claude Le Page wrote: >> >>>Hi Sandro! >> >>Hi Claude, >> >>>For what we are studying , matches or tournaments , I don't see the difference : >> >>I do. >> >>>There are GAMES , and that is all! >> >>So to you tournament and matches for humans are only games? >>So when you play against an opponent you choose random? >> >>>My Thesis is the following : Commercial opening books are supposed to be tuned >>>to the engine with which they are supplied >> >>Some of them yes, others no. >>Of course how they are tuned can be very different from maker to maker. I mean >>if a make a book for program a and I chose a specific line does not mean it is >>tuned...only that I have make a selection...:-) >> >>>Yet they do no better than books of databases , not specially made for an engine >>>or another ; >> >>This is your opinion. Not mine. >> >>>Then there are only 2 possible explanations: >>>1) By chance the books of MegaCor3 and FritzPowerbook happen to be as good >>>opening books as specially tuned ones both for Hiarcs9 and for DeepJunior8 >> >>I do not agree. >> >>>If it's true , then you can use these 2 books as opening book for a new engine >>>that you are developping ; this is amazing , but interesting and useful >> >>My tests show otherwise. >> >>>2)The opening books of Fritz , Junior ,Shredder and so on are obtained by the >>>same way :they choose a database >> >>No, we did not choose a database for Shredder, but made the selection manually. >> >>, and give it's book as opening book for an >>>engine : in this way , I have been said that the opening book of ChessTiger15 is >>>very similar to the book of FritzPowerbook , and yet it is not worse than >>>others >>>There is a point to check:MegaCor3 and Powerbook are "selections" of games ; >>>there is another sort of databases : those obtained in an "automatic" way : >> >>Yes, it is true for ChessTiger and FritzPowerbook, but not for Fritz 8 and Deep >>Fritz 8 as well as for Junior 8 and maybe Junior 9 too...I do not know as they >>are made by other people... >> >>>I have 2 : >>>The games of Chess Informant from N°01 to N°87 >>>The most recent samples of TWIC (only 37000 games) >>>First is perhaps biased : correspondents send the games they think noticeable , >>>so there are more games won by White >>>Second would be more neutral as it contains "integral" of a tornament if it's >>>strong enough (Linares ,Corus aso ), but it's probably less creative than >>>Harding's >>>I am going to test this new sort of books >> >>Look I have all the Chess Informator from 1 to 91 and I did checked many of the >>games listed there. >>The problem with these games is that often the analysis are not correct and >>those lines can be improved. >>Since these games are well known it happens frequently that you play a "winning" >>line taken from those and you get killed by improvements...this is why to get >>games from a database without checking/studying them is time wasted...if it was >>so easy to make a book by getting games from played ones, then I would have >>started to develop a chess program myself rather than working on the opening >>book full (free) time... >> >>>As you can see ,I make no judgement whatsoever >> >>Me too... >> >>>Friendly Yours >>>Claude Le Page >> >>Sandro > >Sandro - Hi, well, cannot give away too many secrets, but can give you some indications: A good book is like a dress specifically made for one person; it should fit perfectly. So, it has to be designed specifically for that player (program). Let's see a good commercial book: 1. Should be made to improve the engine by bringing it to suitable positions. 2. Should include a lot of variety not to bore people. 3. Should not be a copy of ECO or other openings chess books. I mean the variations and frequency of them should be different including also new moves. I mean a sort of updated ECO. 4. Should allow the program to have a specif openings style to make it fun to play with. 5. Should include replies on weak moves too. All this should bring a plus on a chess program. > >can you describe an experimental setup which will demonstrate the value of a >good book? > >I have Shredders 7, 8 & 9 (UCI), Junior 8 & 9, and all the Fritzes going back to >5.32. > >Vas I hope you agree. Sandro
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.