Author: Sandro Necchi
Date: 10:38:07 02/08/03
Go up one level in this thread
On February 08, 2003 at 06:36:04, Uri Blass wrote: >On February 08, 2003 at 05:29:57, Sandro Necchi wrote: > >>On February 05, 2003 at 15:04:52, Chessfun wrote: >> >>>On February 05, 2003 at 14:36:01, Drexel,Michael wrote: >>> >>>>On February 05, 2003 at 14:21:11, Harald Faber wrote: >>>> >>>>>On February 05, 2003 at 10:09:03, Drexel,Michael wrote: >>>>> >>>>>>On February 05, 2003 at 09:32:50, Harald Faber wrote: >>>>>> >>>>>>>It seems to me that this result - and certainly some further results - should >>>>>>>calm many people claiming that Shredder 7 is the strongest program or stronger >>>>>>>than DeepFritz7+Fritz8. But they will never learn that SSDF testing on 2 PCs >>>>>>>with pb-on and tournament time control is somewhat different than betatesting on >>>>>>>one machine with g/60 or even less. >>>>>>>So I admit that I feel confirmation for my sceptical attitude towards >>>>>>>sensational results and reports which arise here frequently about several >>>>>>>programs. :-) >>>>>>> >>>>>>>Harald >>>>>>>(Who cannot remember having seen miracles like new program/version topping the >>>>>>>list with 50pts or more over the former leader) >>>>>> >>>>>>SSDF is testing with slow Athlon 1200 computers. Shredder is known as a slow >>>>>>searcher and he is still relative slow. Shredder 7 lacks tactical ability. >>>>>>Shredder 7 should do much better on fast machines. >>>>> >>>>> >>>>>This is a very old tale I cannot hear anymore. It has never been proven right >>>>>and is always used as an excuse when some engine does not perform as some fans >>>>>hope/expect. >>>>>BTW, what is a "fast" machine for you? P4-2.5GHz? Do you know how much such a >>>>>P-4 is faster than the A-1200? I tell you it is not as much as it might suggest >>>>>if you only look at the numbers. It is even less than the advantage points for >>>>>faster hardware could compensate the 14-22. >>>>> >>>> >>>>I dont think im a Shredder fan. I have Shredder 7, Fritz 7,8 and a lot of old >>>>commercial chess programms plus some old chess computers. >>>> >>>>No, a P4 2.5 GHz is not a fast machine for CC. A dual MP 1900+ or 2000+ is. >>>>of course Deep Fritz 7 would be stronger too, but I think it is more >>>>important for Shredder 7 to play on a very fast machine. >>>>btw, as you know 14-22 doesnt mean much. 40 games per match (SSDF) is IMO not >>>>enough because chance (be lucky) plays an important role in CC. >>> >>> >>>Sandro made no mention of needing a faster CPU than a A1200 when he posted here >>>that Shredder 7 will be 50 points (I think it was) stronger than Shredder 6. And >>>further that it will top the SSDF. >> >>Dear Sara, >> >>yes you are correct, BUT here is the complete story. >> >>1. Shredder 7 is stronger but has problems in the middle game when facing Deep >>Fritz and Fritz 8 due to the different evaluation of the bishops and knights >>mainly. >>This makes very important the openings which are selected. If Shredder gets a >>weaker position from the opening very hardly will win the game while with an >>advantage sometimes it does sometimes it does not. >> >>2. We have told SSDF to test the UCI version and they did not. I am not going to >>say anything anymore against SSDF because they simply do whatever they like. I >>do not understand why they ask if they do the opposite. >> >>3. Theorically the UCI interface and the CB are with the same book, but >>practically not. UCI book has 3 options: Tournament, standard and wide. CB only >>2: standard (called tournament)and wide. The tournament option in UCI is more >>restricted playing only the a to c moves while the standard play also d moves >>even with a low frequency. >> >>The reason why the book is so wide in standard mode is because I have been asked >>to do it this way. It is good because the variety is quite higher, but by >>increasing the alternatives one increases also the weak variations and moves. >>This means to loose games expecially against a strong program like Deep Fritz 7 >>or Fritz 8 where the program is not able to recover a weak position. >> >>4. I have made a stronger engine in the middle game which should be scoring >>better against Fritz 8 and deep fritz 7. It is called Shredder El Cid, however >>it seems that against other programs may turns out to be weaker. This means that >>to find out the real strenght of a program it is necessary to test it against >>many different opponents and not make conclusions agains one specific opponent. >>On such a large opponents the program should turn out to be about 50 points >>stronger. >> >>I hope this clearify the matter. Hi Uri, > >I can add that from my experience fritz8 did not play well in the only game that >I tested it against Shredder7 at long time control. Well, there are positions which are just the opposite. It really depends on the openings. Of course increasing the time the computer play becomes more precise so weak moves usually go badly. Maybe your novelty was not that strong. > >I used it at level of 12 hours/40 moves to test a novelty that was suggested by >Fritz8 after a long analysis(the novelty was not suggested directly but only in >the 3th move of the pv of Fritz when the first move of the pv is already a move >that is not in shredder book but shredder has reply with white). > >Fritz8 started with evaluation that was close to 0.00 but it's evaluation went >down and I stopped the game in a position that it seems that shredder should >win(shredder said more than +3 for itself and Fritz8 said almost +2 against >itself). In these positions the chances that they are both wrong are about less than 5%. > >Fritz played for my side and the result of Fritz's poor moves is that I decided >not to go for the line with the novelty but for a line that was played by GM's >in the last years. > >I am not sure if the novelty is so bad and Fritz8 simply played with no plan but >I decided that I have not time to test it more seriously. > >Both sides used 64 Mbytes hash on A1000 Mhz and the result was that both sides >were slowed down in the first minute of the search but later ran normally. > >Maybe I should use smaller hash tables but I am not sure about it and >practically the engines ran at normal speed in more than 90% of the time that >they calculated. It seems most programs do not play the best using more hash tables. Shredder seems quite optimized also with a lot of memory. > >I do not post the game now because I do not know if it is right to post >correspondence games before they are finished inspite of the fact that I do not >think that it can help my opponent because I already played a move to get out of >the Shredder7-Fritz8 game. Yes, I agree. > >Uri Ciao 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.