Author: Sandro Necchi
Date: 22:57:54 02/18/06
Go up one level in this thread
On February 18, 2006 at 16:38:58, Uri Blass wrote:
>On February 18, 2006 at 15:20:40, Sandro Necchi wrote:
>
>>On February 18, 2006 at 14:32:40, Drexel,Michael wrote:
>>
>>>On February 18, 2006 at 13:18:26, George Tsavdaris wrote:
>>>
>>>>On February 18, 2006 at 07:07:43, Drexel,Michael wrote:
>>>>
>>>>>On February 18, 2006 at 05:51:45, Sandro Necchi wrote:
>>>>>
>>>>>>
>>>>>>I am expecting Rybka to get a SSDF list Elo about 150 points above Shredder 9,
>>>>>>but in May/June it will not be no. 1 anymore.
>>>>>>
>>>>>
>>>>>Yet another ridiculous expectation of yours.
>>>>>We're getting used to it.
>>
>>Michael,
>>
>>let's comments your claims:
>>
>>>>
>>>>He is expecting 2 things:
>>>>1)Rybka will be 150 points above Shredder 9.
>>
>> THE SSDF RATING LIST 2006-02-10 %105197 games played by 275 computers
>> Rating + - Games Won Oppo
>> ------ --- --- ----- --- ----
>> 1 Fruit 2.2.1 256MB Athlon 1200 MHz 2853 26 -25 835 71% 2701
>> 2 Hiarcs 10 HypMod 256MB Athlon 1200 MHz 2845 33 -31 559 75% 2652
>> 3 Shredder 9.0 UCI 256MB Athlon 1200 MHz 2815 23 -22 1035 69% 2675
>>
>>I am claiming that Rybka 1.2 should be about 2965 points when tested.
>>
>>My claim is based on the following:
>>
>>1. Rybka version 1.02 beta was about 100 points stronger than Shredder 9.
>>2. Rybka version 1.13d is stronger, maybe some 40 points.
>>3. Rybka will be improved in the endgame starting from version 1.14
>>4. Rybka version 1.2 will have a strong opening book specifically made for it by
>>one of the best book makers (Jeroen Nooman; I hope the spelling is correct)
>>
>>all this togethe should score about 150 points above Shredder 9 in the SSDF
>>list.
>>I use "about" because I am too busy to make specific tests and it insteat of 150
>>it is 140 or 160 it would not be that different.
>>
>>I am not the only one to expect this.
>>
>>P.N. "Expect" means that for me this is achivable, but I am not making it so it
>>depends on some one else work...do you understand what I mean?
>>
>>this is the latest
>>
>>>>2)Rybka will not be no1 in May/June.
>>
>>Yes, I am expecting some program to improve a lot thanks to weaknesses shown by
>>Rybka which will be fixed and using the experience made by these programmers
>>which are in the field since several years they will inprove their programs a
>>lot.
>>
>>P.N. I said programmers and not "programmer" so I am not referring to SMK only.
>>
>>>>
>>>>In what expectation are you refering as ridiculous?
>>>
>>>Both, I seriously doubt Rybka will come close to 2965 at SSDF and
>>
>>Well, if you would make tests 24 hours a day like me and having all the chess
>>programs which are available you would be a lot less doubtful...
>>
>>>I understand 2) as "Shredder 10 will be rated more than 150 points higher than
>>>Shredder 9 in June".
>>
>>I did not state this.
>>
>>I said that another program will be no. 1 without mentioning Shredder...you
>>understood it that way...read my post again.
>>
>>I believe SMK can do it, but this is based on my evaluation of his ability and
>>knowledge as well as on the time he can spend on this.
>>I am not sure the time he can spend will be enough, yet.
>>Still I am expecting other programmers to improve a lot too (if they want and
>>can spend enough time working on the engine).
>>
>>>
>>>> Also by "yet", i conclude that you think he made in the past one or more
>>>>ridiculous guesses. Can you give an example.....?
>>>
>>>He predicted several times the next version of Shredder will improve
>>>by at least 50 points at SSDF.
>>
>>This is a true lie.
>>Uri has stored all my claims on this and can tell exactly what I claimed.
>>
>>Anyway I report here what I said to refresh your memory:
>>
>>1. I said that I did not want to make any claim for Shredder 9 as people were
>>going to attack me even for 1 or 2 points less. I said that I had no time to
>>make a lot of tests to check correct info because I was too busy working on the
>>opening book, so I could only give indications/expectations and I was expecting
>>Shredder 9 to score about 30 points more than Shredder 8.
>>No the difference is 9 points, but Shredder 9 has been the program against which
>>all the others tested mainly and this means that at this point is clearly
>>handicapped in the score, so 9 more is quite good still as it should have been
>>much more lowered, but here the UCI gui did help to keep it ahead of Shredder 8
>
>
>Some questions:
>1)Is shredder9 30 elo better than shredder8 if you do not count programs that
>were developed before shredder9?
THE SSDF RATING LIST 2004-04-22 97872 games played by 264 computers
Rating + - Games Won Oppo
------ --- --- ----- --- ----
1 Shredder 8.0 CB 256MB Athlon 1200 MHz 2818 34 -32 481 70% 2673
2 Shredder 7.04 UCI 256MB Athlon 1200 MHz 2809 24 -23 967 71% 2648
3 Deep Fritz 8.0 256MB Athlon 1200 MHz 2790 26 -25 855 72% 2625
THE SSDF RATING LIST 2005-07-29 %102623 games played by 272 computers
Rating + - Games Won Oppo
------ --- --- ----- --- ----
1 Shredder 9.0 UCI 256MB Athlon 1200 MHz 2821 28 -27 704 67% 2697
2 Shredder 8.0 CB 256MB Athlon 1200 MHz 2805 23 -22 1115 71% 2648
3 Shredder 7.04 UCI 256MB Athlon 1200 MHz 2804 22 -21 1133 69% 2663
The first SSDF list showed this.
Note also that the latest previous program is handicapped when the new ones come
as it is clear that testing has been mostly against the previous best one, so we
need to look the first rating to check the "guessed" improvement.
The first list showed 16 points and I think if went up to 20 or 22 points in the
next one or two list. Since I said that I would guess about 30 points I was not
too far. AGAIN I NEED TO REMEMBER THAT I SAID MINE WERE INDICATIVE VALUES AS I
DID NOT HAVE THE TIME TO MAKE A LOT OF TESTS TO VERIFY THIS CORRECTLY.
The reason of less scoring than expectation is that Shredder 9 is less
aggressive and did not score as it should against some of the weaker
opponents...too many draws...
>
>2)You admit that UCI gui helped shredder9.
>Does it mean that the situation is even worse for shredder9 and maybe there was
>no improvement in the engine from shredder8 to shredder9(UCI interface is not
>an improvement in the engine and the uci interface can be used also for
>shredder8).
According to my tests the GUI helps as follows:
1. Better learning, so this may is worth 7-10 points
2. Better book (if tournament mode is selected; more selective). This should be
worth about 15 points.
considering that the gain was about 22 points, we should say that there was no
real overall improvement due to not enough wins against weaker competitors. Also
we lost the match against Junior 9 suffering the aggressivity of that program.
However, when I was telling that I was expecting 30 points more I was referring
to Shredder 8 CB as this one was the one tested by CB, so no RIDICOLOUS
expectations!
>
>Uri
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.