Author: Rajen Gupta
Date: 02:38:51 06/09/01
Go up one level in this thread
On June 09, 2001 at 02:27:27, Uri Blass wrote: >On June 08, 2001 at 18:04:52, Rajen Gupta wrote: > >>On June 08, 2001 at 10:44:29, Harald Faber wrote: >> >>>On June 07, 2001 at 11:35:16, Rajen Gupta wrote: >>> >>>>On June 07, 2001 at 09:53:24, Harald Faber wrote: >>>> >>>>>Clear win for Tiger, no chance for the oldie. 7.5-2.5 in favour of Tiger. Even >>>>>booklines up to 30 moves (10th game) didn't help...games, as always, at >>>>>www.geocities.comHarald1312/HaraldFaberE.html. >>>> >>>>hi hrald: how are you coducting this match-on 2 computers/ and what opening >>>>books are you usinig? >>> >>>I am using the according opening books, that means the normal ct.tbk for Tiger >>>14 and the tournament book is activated in MChessPro 8. >>> >>>>i am running a match on a single celeron 900, eng vs eng, 64 mb hash tables >>>>each; 90min/60 moves; both programmes using an individual copy of the chessbase >>>>tiger book. currently chessbasetiger14 vs junior5 w=10; l=7 d=10. at 20 games >>>>both were running equal with w=6, l=6 d=8. i'm not sure if the latest progs are >>>>that much stronger if the oldies can play of the same opening books. >>>> >>>>rajen >>> >>>Don't forget that many opening books are adjusted to the engine, so testing with >>>only one book to which all programs have access just lets you test this book. If >>>you want to check a program, you have to use the given opening book. That is why >>>I am not very lucky about the fact that the Chessbase Tiger does not have the >>>Noomen-opening book. I am sure that Tiger is strong enough to play well with >>>this opening book too, but I am not sure if the results would be better with the >>>Noomen-book. >> >>i'm using the tiger book for both programmes and the engines are running in the >>tiger interface. (so no excuses for the tiger)assuming that the chessbase tiger >>book is not optimised for tiger(which i doubt very much-in my opinion the >>chessbase book is at least as good as the noomen book,if not better)it certainly >>is not optimised for junior 5. i initially thought of using the junior5 book for >>both programmes but then there would be the excuse that tiger is using an >>un-optimised book. the fact that as of today the score is tiger w=14;l=11; d=15 >>shows that there is very little to separate tired old junior5 from the so called >>'monster" chesstiger. >> i am a firm believer in the theory that 99% of the so-called improvement in >>newer programmes comes from the opening book and that basically any good opening >>book consisting of high level gm games will equally suit all programmes >> >>rajen > >99% of the difference? > >The result suggest 30 elo difference between Junior5 and tiger14. >if 99% of the difference between Junior5 and tiger14 is because of better >opening books then it suggest that the difference between tiger14 and Junior5 is >3000 elo. > >Another points to consider: >1)30 elo can be wrong and the statistical error is still more than 20 elo(you >need hundreds of game in order to have statistical difference of 20 elo) > >2)The match is done on one computer and it is possible that tiger14 is better on >2 computers(I do not think that it create big differences but the difference can >be something like 20 elo) > >3)You cannot decide which program is better by match that is only between 2 >programs because it is possible that A beats B, B beats C and C beats A. > >Even if tiger14 cannot get significant result against Junior5 you need to >compare also the result of Junior5 and tiger14 against other programs. > >4)It is possible that Junior5 likes your hardware more than tiger14(if Junior5 >is 2 times faster relative to the ssdf hardware when tiger14 is only 1.7 times >faster relative to the ssdf hardware it can explain part of the difference) > >Uri hi uri:the match is done on 1 computer but neither of the 2 engines are losing any cpu time when they are calculating.i have compared this with the figures on the "tigermark" and "juniormark". if tiger likes 2 computers better, so would junior, and i think equally. i am conducting a long series of matches-probably 100 games, so i would get a better significance. but i dont see much difference-whenever tiger wins a game, juniuor wins the next one and the scores have remained consistent.i dont expect more than a 6 game difference at 100 games. i will compare the relative performance against other programmes as well;but again i think the score ould be extremely close esp when the opening books are standard. rajen
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.