Author: Robert Hyatt
Date: 13:35:47 09/26/00
Go up one level in this thread
On September 26, 2000 at 05:23:07, Uri Blass wrote: >On September 25, 2000 at 21:33:28, Vincent Diepeveen wrote: > >>On September 25, 2000 at 09:22:30, pavel wrote: >> >>>On September 25, 2000 at 08:53:37, Vincent Diepeveen wrote: >>> >>>>On September 24, 2000 at 22:56:34, Mike S. wrote: >>>> >>>>>On September 24, 2000 at 22:34:50, Jorge Pichard wrote: >>>>> >>>>>>Of course that my second choice would have to be the King an experimental >>>>>>version of the future CM8000. >>>>>> >>>>>>http://ourworld.compuserve.com/homepages/thstorm/partic00.htm >>>>> >>>>>I wonder why Quest (=Fritz) should run on a dual Celeron 433 only? >>>> >>>>He used dual 433 previous year. >>>> >>>>>Btw., Nimzo was my favourite for the London WCh. I predicted the outcome (but >>>>>the participants didn't stick to my prediction): >>>> >>>>>http://f23.parsimony.net/forum50826/messages/515.htm >>>>> >>>>>For the Dutch Open, I think that - in addition to the 4 professionals - Diep >>>>>(dual PIII/800), furthermore a new version of Kallisto II, and probably Patzer, >>>>>are the dark horses which deserve attention. >>>> >>>>There are 2 programs dual at dutch open: Diep and Quest. >>>> >>>>However let's face it that there are 3 programs out there with a real >>>>good and especially well tested book: Nimzo and Quest and Tiger. >>>>From those only a single program is dual. Nimzo heavily relies upon >>>>outsearching its opponents after having an openingsposition that's better >>>>for it. >>>> >>>>The King has a book which is not that bad, but definitely not anymore >>>>a 'todays' book. Further The King is an engine which plays very interesting >>>>chess at slow hardware. Just sacrafice a pawn and play a cool game. At >>>>nowadays programs searching 11 ply or more with very aggressive tunings, >>>>that simply is getting a more hard way to play chess, as nowadays they don't >>>>only swallow the pawn, but also win the game. >>>> >>>>apart from that, i doubt whether The King will *ever* win from DIEP, >>>>as from testgames it appears that DIEP is the worst opponent for The King, >>>>as DIEP is not only outsearching The King positionally (not tactical), >>>>but also having more knowledge, so the only advantage the king has against >>>>DIEP is better tactics, which nowadays hardly works. This where nimzo's >>>>cool tactics work usually excellent as it's backupped by a superb book. >>> >>>you kidding me !! >>> >>>can you paste some of the sample games? >> >>Yeah plenty, just go to a chessserver and look to cm7000/6000 and others >>playing diep. Note past years at dutch open The King didn't manage to win, >>all games for the same reason. >> >>>*ever* win is (more than) a bit harsh IMO, you can come to such conclusion >>>because DIEP is not as much commercially available as the king (CM) engine is. >>>So only you and few others get the privilege of testing DIEP with CM. >> >>This is not true. I don't test a single game against CM by hand against diep, >>i only see games as played on the internet and at dutch open. >> >>CM doesn't allow to play unattended somehow as far as i know. >> >>>by the way I think DIEP has played more tournaments than any other chess program >> >>this is not exactly true. perhaps i play the tournaments you take a look at! >> >>i join basically world champs, dutch champ and german champ. Other tournaments >>i usually depend upon whether people want to operate it, though if spain >>champ would get organized at a better date i would go there too. >> >>Of course it's true that i join even if it's sure that i have just introduced >>a big number of bugs in the program, like for wmccc. Also it would be result >>wise perhaps not so smart to join dutch open. despite that i fixed many >>of the bugs as introduced before wmccc, i'm just not ready for dutch open >>to face the Kure and Noomen books. >> >>>I know of......so how many did it win? (not that it is of any importance) >> >>I won paderborn nearly 2 years ago, after diep fought back from all lost >>positions. Like against shredder it was a pawn down, but got to a won >>endgame, but then blundered the endgame. It was not so well out of book >>against nimzo, but got also a won far endgame, but blew it to a draw too, >>despite the loss also in endgame against P.Conners, it also could have >>gotten a draw there with a better endgame. >> >>It has a better endgame now and still improving, but in the meantime the >>commercial books have gotten really better. Shredder is already having >>problems keeping up with the new books, but with superb endgame play it >>still manages to keep draws in lost positions resulting in a worldtitle. >> >>Note that diep lost end of 99 also the spain champ title by losing to >>tiger in the one last round. It appeared that this was also a lost bookline. >> >>I was using some rather old openingsbooks (from 20 years ago) and some >>lines are completely refuted. I had prepared the line that came on the >>board, but it is nowadays seen as won for black. >> >>So it's obvious what i and MANY others should focus at right now: improve >>book. >> >>Like nimzo in wmccc 2000 came out of book against SOS with +mate_in_12. >>12 moves later... >> >>Zchess came out of book against nimzo with over a pawn down, lost the >>pawn and the game real soon. >> >>Most engines aren't far from the strength from nimzo. >> >>In fact i don't have much respect for nimzo as an engine. It's just >>an aggressive tuned engine which is tactical real strong, good blitz program, >>excellent in doing nothing. But compared to most other engines it is >>much worse. >> >>I'm amazed people keep on betting on this engine. >> >>I find my own program, zchess, SOS, The King, Gandalf and many others >>miles better as Nimzo. >> >>Yet i doubt whether The King's book will ever get to equal standards with >>nimzo. >> >>As long as nimzo can win games with a +mate in 12 score on its screen, >>then it can still go for tournament wins. >> >>Yet even with a piece up out of book against shredder it couldn't win! >> >>With a pawn up against diep short after book (pawn win was forced by >>book) in dutch open 1998 it couldn't win against DIEP. >> >>I'm amazed people bet on nimzo. It's made to solve testsets and to >>finish games from a won position. Obviously with material up after book >>you *can* chose for this approach, and i'll never blame Chrilly for that. >>he has made something that sells, simple as that! >> >>In the end most people are not >>very in depth studying programs. They just care for the number of points >>it scores and whether it solves a few tactical shots in testpositions which >>all programmers saw already years before the user saw it. >> >>If i manage to not only get a better book, but also test in which lines >>diep plays well, then i'll bet it will do real well too. Right now i'm >>loaded with other work, so i can't do that within 3 weeks time. >> >>Zchess book will also improve bigtime. I'm sure Insomniac too. >> >>I think James Robertson also was bigtime dissappointed in WMCCC about how >>many games were won just on book. People just don't realize it till they >>join themselves in a world champ, or when they analyze games of it. >> >>I would have had 2 titles at least with a better book with DIEP and with >>the current endgame of it. Zchess would have perhaps already had a world title >>if jaap v/d herik had done a fair pairing, and nimzo would not >>get close to winning any tournament if people would have a better book. >> >>Obviously it will be only a matter of time before books improve of the >>different programs. Mine for sure will. Zchess for sure will. >> >>The King i doubt. Its book was never real bad, but it's simply not getting >>free points like nimzo does. Shredders book is not bad, but definitely >>a mile behind Kure/Noomen. >> >>Most engines will remain losing points because of it, yet a number of them >>will improve this. >> >>Getting out of book with +mate in 12 is just unbelievable. >> >>Note that Kure book to human standards still sucks. A human is MUCH better >>prepared as any todays playing engine is. >> >>Of course with exception of SSDF testing, as you test against something >>from which you know what it is gonna play. I'm relating here to tournaments. >> >>I'll be really amazed at dutch open if nimzo doesn't get a few games with >>over +2.xx out of book. > >This is something that I do not understand. >It is easy to prevent coming with -2.xx out of book by having a small book that >every move in the book was analyzed manually and playing for quiet lines. > >Sos played for a tactical line with mate attack out of book. >If you add the fact that Sos is weak in tactics and could not see a simple mate >you can understand the fact that nimzo had evaluation of mate out of book. > > >Shredder blundered against nimzo with Qxb2. > >I believe that Shredder could avoid this mistake by using better strategy when >database say 100% and only one move for the opponent. > >It was easy to guess without book when you use a big database that the move of >nimzo is going to be Bxb4(100% score and the only move out of book) so shredder >could instead of playing Qxb2 to wait and calculate the position after Qxb2 Bxb4 >and after seeing the drop in the evaluation to go back and find a better >move(Rxe1+ instead of Qxb2). > >If there is not a drop in the evaluation it can save time by playing the next >move in 0 seconds and the only cases that this strategy cause loss of time in >the clock are cases when the mistake was a previous move or cases when the >opponent does not play the known move that gave 100% result from human-human >games. > >Uri This only partially works. IE I have such a mode in Crafty (book random 0). The problem is that many of the book lines have traps deep enough that a search won't see the problem. Of course, it will see some of them, and avoiding even one bad line is a good idea. But it doesn't cure the problem 100% of the time.
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.