Author: Dirk Frickenschmidt
Date: 10:56:20 09/13/98
Go up one level in this thread
On September 13, 1998 at 12:55:08, Bruce Moreland wrote: Hi Bruce, > >On September 13, 1998 at 12:07:14, Thorsten Czub wrote: > >>what an amazing disadvantage to use the book that was sold for extra >>money to make sure fritz scores better, and that was used in xyz % >>in sweden and that fritz-experts call important and at least >>giving 50 elo points in usage. >> >>What an amazing disadvantage to have a book that is 530 Mbyte big and >>comes especially to help fritz... > >What I am trying to steer these threads around to is a discussion of big, >automatically generated books, and how to implement blunder-avoidance in these. > >If I was trying to preen for the SSDF, and I had a lot of resources, I would >make the huge book, autoplay against my competitors, and trim out lines that led >to obvious bad situations. > >This would leave a huge book with some of the crap removed. > >If you play against other opponents, you get new crap though, since the other >opponents play different lines. > >This is what I've noticed with my large book. I feel pretty comfortable that I >will come out of book OK against some opponents, because when I see a bad line I >trim it (this can happen either against a human or a computer), but when I play >a new opponent I see some worse book lines. > >If I never played any new opponents I would say that my book was working great. > >Maybe Fritz is experiencing the same problems. I had the same impression, just like you trying to find a technical answer concerning Fritz' bad opening choices. Instead of any cheating suspicion and really not touched by how good or bad Frtitz5 is doing in this tournament, I just was interested how moves, which perhaps had a low evaluation in the original powerbook, now appeared more frequently and disadvantagous. I'm just interested to know how it happens and how it perhaps can be avoided. Therefore I had written: --Begin of quote-- I have not read from their post that they accuse you of cheating. I wouldn't either. Concerning the Fritz powerbook: 1. It definitely *has* weak lines in it. I have seen one from time to time as well. Probably this comes from having too many unchecked lines from below strong GM level in it. 2. With strong learning on, if you play some blitz games or games against different or weak players, the narrow learning sometimes specializes in a funny way, resulting in unrealistic weights for some moves which are no good any longer for serious play. 3. To be sure that no such previous kind of bad-tuning by random play happens, I clean the evaluations before serious tournaments. Just to be sure I get the standard book lines and learning only from the serious tournament games played then. I don't know in which state your powerbook was when beginning the tournament. 4. I said I get these funny openings in the powerbook as well. But what perhaps looks strange to some here, is *how* *much* bad luck in finding such openings Fritz5 had so far. I do not evaluate this in any way, not knowing at all how it happened. I have not rechecked all of the games and only replayed some that where of interest to me, but I remember the last one was not the first bad Fritz opening choice in this tournament. 5. My technical proposal is just to recheck what I said above in 2. and 3. If none of that is the reason for the bad opening choices, well then its no drama either. Fritz then just made bad powerbook choices, resulting in some bad middlegames, on a processor on which it is relatively slow compared to other programs anyway. No big deal... There are enough interesting programs and games in your tournament, so I for one won't stare at the Fritz5 outcome all the time... --End of quote-- I want to clarify that I have never and would never imply any cheating from Thorsten's side, as is obvious in my post above, though Thorsten repeatedly claimed such attacks including my name. Kind regards from Dirk > >If I was a chess-playing customer of Fritz I might want the large book, just so >the diversity of openings played is increased. You have been messng with these >programs for years, so you know exactly what I mean here. You can go all the >way back to Chess Challenger 7, which was out of book after d4 Nf6, or you can >come forward to MChess 3.5, which as far as I remember always played the >Sicilian Dragon if you let it, or you can come forward to the most modern >programs, which still have narrow books which let you play the same lines over >and over. I don't play a lot of chess now, but I can tell you that I would have >loved to have had lots of opening book diversity when I was playing a lot of >chess. > >bruce
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.