Author: Sandro Necchi
Date: 08:22:53 12/09/02
Go up one level in this thread
On December 09, 2002 at 09:58:13, Uri Blass wrote: >On December 09, 2002 at 09:27:49, Sandro Necchi wrote: > >>On December 09, 2002 at 09:08:54, Uri Blass wrote: >> >>>On December 09, 2002 at 08:17:37, Sandro Necchi wrote: >>> >>>>On December 09, 2002 at 07:27:51, Uri Blass wrote: >>>> >>>>>On December 09, 2002 at 06:44:55, Harald Faber wrote: >>>>> >>>>>>On December 09, 2002 at 05:15:08, Jouni Uski wrote: >>>>>> >>>>>>>To get REALLY improved top engine, I mean! >>>>>>> >>>>>>>According to SSDF Deep Fritz 7 is 14 points stronger than Fritz 7 (and probably >>>>>>>Fritz 8 is very close) >>>>>> >>>>>> >>>>>>IMO DF7 is more than 14pts stronger than F7. >>>>>> >>>>>> >>>>>>>and Tiger 15 is 16 points over Tiger 14. Not so much. >>>>>> >>>>>> >>>>>>Needs further testing. >>>>>> >>>>>> >>>>>>>I quess, that Shredder 7 will have much more progress. >>>>>> >>>>>> >>>>>>That is based on marketing placed by Sandro and Helmut. Simply a GUESS. Maybe >>>>>>you should also try to GUESS how many former engines have also been announced to >>>>>>be 50pts or more stronger than the predecessor and FAILED. >>>> >>>> >>>>Pls. note: I DO NOT AND NEVER MADE MARKETING! >>>> >>>>>>Believe me, much more than one. Actually, as it seems, Tiger 15 falls into this >>>>>>category. So calm down, wait and see whether Shredder 7 will really show an >>>>>>improvement of 50pts. >>>> >>>>We have made a huge amount of games and based on those we can state that the >>>>most reliable figure is 50 points. >>>> >>>>Pls. consider that the opening books have a huge no. of variations and depending >>>>on them and the selection toward the opponents this can vary, THIS IS WHY I AM >>>>TALKING ABOUT MOST RELIABLE GUESS! >>>> >>>>>> >>>>>> >>>>>>>Note, that Shredder 6 is 55 points over 5.32 and Paderborn version full 72 points. Do it Stefan! >>>>>> >>>>>> >>>>>>Did you notice the difference between DeepFritz 7 K6-450 <-> A-1200 and Tiger 15 >>>>>>K6-450 <-> A-1200? What do you say now? >>>>> >>>>>I do not know and I guess that tiger was simply unlucky on A1200 and lucky on >>>>>K6-450(there is a statistical error) >>>> >>>>Yes, but I was talking about 45 points improvement. Pls. note program >>>>improvements so no link with a specific hardware. Also the program may was a >>>>little lucky with slow hardware so more or less this is the improvement. >>>>Why Uri does not admit I was right rather than criticizing every time he think >>>>he has the opportunity? >>> >>>It is not something that I have against you. >>>I simply know from history that the estimates for shredder were too optimistic. >> >>Well, it happens ones and you are repeting that all the time as a rule. Note >>that when that happen I CLEARLY SAID THAT THIS WAS SOMETHING TOLD ME BY A TESTER >>FRIEND OF MINE... >>If who was selling Shredder did so, that has nothing to do with my claims. AM I >>WRONG TO STATE THIS? >> >>Pls. say it once forever! >> >>> >>>> >>>>Again I repeat that I do not care about real Elo improvements as my interest is >>>>to improve my opening book and win the WCCC! >>>> >>>>> >>>>> >>>>>Note that the predictions were also better at the time of shredder6 >>>>> >>>>>Here is a link to the prediction >>>>> >>>>>http://ccc.it.ro/search/ccc.php?art_id=199339 >>>> >>>>Pls. remember that I was talking about 50 points improvement based on UCI >>>>version; the one with my book, which was not tested by SSDF on A1200. >>>> >>>>> >>>>>The prediction were at least 80 elo better for shredder6. >>>>>The prediction was also for sandro's book but I understand that shredder >>>>>passerborn is using that book. >>>> >>>>Yes, it does, but using the book made for Shredder 6.0, not for this different >>>>version, so not fully optimized! >>> >>>I understand your explanation but I expect better engine with the same book to >>>perform better. >>> >>>If shredder padderborn is weaker than shredder6 with the same book when your >>>book is used then it seems that in some meaning it is weaker than shredder6. >> >>This is what I claimed: >> >>Shredder 6 (UCI as referred with my book; Shredder 7 will have my book on both >>GUI, but this will be the first time). better about 80 points in total: >> >>40-50 points the engine and about 15-20 points the book and the rest the new >>learning (again UCI GUI). >> >>Shredder Paderborn: about 15-20 points stronger (the engine). >> >>So now since the book was made for the 6.0 which it is REASONABLE that >>Padderborn version was not optimized for the book as usually it is the other way >>around! >> >>So my statements are turned out to be REALISTIC considering what above WHICH IS >>OBVIOUS TO ME and should be to all if you think about it! >> >>Did I ever critized another program? No I did not. >>Did I ever talked about other programs? No I did not. >>Why do I give info on Shredder? Because being in the developing team I think it >>is an enrichement for this forum to know things from the people that work on >>them. When I was a customer I wanted this, but very seldom they were available. >>Why other people, which work on top programs, do not appear on this forum? >>Because there are too many people TOO READY TO CRITICIZE and made general >>statement instead of trying to understand things. >> >>I hope I have been MORE than clear. > >I understand > >It seems based on the claim that your book was good for shredder6 and not good >for shredder padderborn that these programs are completely different and there >is a significant change that is productive in part of the cases and counter >productive in another part of the cases(otherwise a book that is good for >shredder6 should be also good for shredderpadderborn). Yes, just one example: If you are an aggressive player it is not worth to force you to play a Carokann defence line where you have to play more than 15 defence moves. You'll be "somehow force" by your aggressive style to force an attacking move and get a bad position. The same is true for chess programs. One small example: My Shredder 6.0 book gives 1. d4 as best move with white. For Shredder Paderborn it should be 1. e4 So what I mean is that Shredder Paderborn is a little stronger than 6.0 (I think around 15-20 points), but because the book is not suiting it's style it may gets 15-25 points penalty from the book resulting a litte less stronger in the SSDF list or about the same. If this is true we may have missed the first place in the list when Deep Fritz 7 was not available yet. So my complains with SSDF people to test the UCI version where not too wrong at the end. Sandro > >Uri
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.