Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Chess Genius: The rodney dangerfield of chess programs?

Author: Jonathan Lee

Date: 10:49:01 10/24/99

Go up one level in this thread


On October 24, 1999 at 08:08:53, Hans Christian Lykke wrote:

>On October 24, 1999 at 00:38:20, Christophe Theron wrote:
>
>>On October 23, 1999 at 14:15:01, blass uri wrote:
>>
>>>On October 23, 1999 at 10:44:54, Jonathan Lee wrote:
>>>
>>>>On October 22, 1999 at 13:45:17, blass uri wrote:
>>>>
>>>>>On October 22, 1999 at 11:27:00, Jonathan Lee wrote:
>>>>>
>>>>>>Chess Genius 6 is about double the cost than Fritz 5.32 and others; that's sad.
>>>>>>
>>>>>>I remember the game show Family Feud, and the asked who are the best geniuses?
>>>>>>They are Einstein, Edison, Da Vinci, (and Alexander Graham Bell?).  I would like
>>>>>>to have Richard Lang to be one of them.
>>>>>>
>>>>>>Yes, Lang has (extremely most probably) changed the landscape of Eastern
>>>>>>European chess.  Russia's official sport relies directly or indirectly on Lang's
>>>>>>tactics.
>>>>>>
>>>>>>Competition and change of computers are rather rapid these days.  There may have
>>>>>>been "reverse engineering" of a dozen software companies to equal Lang's code.
>>>>>>In America, it is hard to copyright any software; after all it is all zeros and
>>>>>>ones.
>>>>>>
>>>>>>If one had FischeRandom or other chess variants, Lang would be the starting
>>>>>>point.  Then, they would add an opening library along the way.
>>>>>>
>>>>>>There will be software improvements on chess code, but the changes won't be as
>>>>>>large as the early 1990's.
>>>>>>Jonathan
>>>>>
>>>>>How do you know?
>>>>>I expect big improvements in Junior,Rebel and tiger.
>>>>>
>>>>>Uri
>>>>The quantum rating leaps are (if told truthfully by their advertisements):
>>>>1. 2265 USCF to 2325 USCF (Fidelity)
>>>>2  2325 USCF to 2424 USCF (Mephisto combines Fidelity for Genius 1)
>>>>3. 2424 USCF to 2470 USCF (Genius 2)
>>>>4. 2470 USCF to 2495 USCF (Genius 3)
>>>>5.?   2500+ USCF (hardware improvements and Genius 5)
>>>>6. Deeper Blue is probably 2700 USCF
>>>>Please note other software companies have finally caught up with the "Lang
>>>>explosion".
>>>>I am going to assume that there is a 75 software rating difference between
>>>>Genius 3 and Fritz 5.32.  I don't want to go negative, but I find hard to
>>>>believe larger improvements are under way;  there will be lots of small ones as
>>>>I said in the previous (6th) message OK.
>>>>Other softwares (word processing, spreadsheets, AUTOCAD drawing, etc.) will also
>>>>have evolutionary (not revolutionary) improvements too.
>>>>(This is my 7th message ever.)
>>>
>>>1)Fidelity,Genius1,Genius2 seems to be the main significant improvement.
>>>(205 elo between before Fidelity to Genius2)
>>>What is the difference in time between them?
>>>What is the difference in ssdf rating(there is not enough information about
>>>games against humans)
>>>
>>>2)The question is about the software improvement so
>>>I want to know if they play on the same hardware.
>>>
>>>3)I think that the difference between Genius3 and tiger12 on the same hardware
>>>may be more than 75 ssdf rating.
>>>Genius3 is not tested on the same hardware as tiger12 but Genius3 was tested on
>>>the same hardware as previous version of tiger and I remember that the
>>>difference was close to 75 elo.
>>
>>
>>I don't see Genius3 in the last SSDF list, but I do see Genius5 on P90. And
>>Tiger 11.8 was tested on P90 too.
>
>This from the latest SSDF-list (long!) Genius 3 on 33.
>
> 23 Chess Tiger 11.8  Pentium 90 MHz        2483   54   178   57%  2435
>  24 Gandalf 3.0  64MB P200 MMX              2468   41   307   59%  2400
>  25 193  64MB P200 MMX                      2465   42   279   44%  2508
>  26 Kallisto II  64MB P200 MMX              2446   35   403   52%  2431
>  27 Rebel 8.0 Pentium 90 MHz                2440   18  1435   55%  2405
>  27 Rebel 9.0 Pentium 90 MHz                2440   23   890   47%  2461
>  29 Hiarcs 6.0 Pentium 90 MHz               2437   18  1437   51%  2433
>  30 Genius 5.0 DOS Pentium 90 MHz           2431   18  1465   48%  2444
>  31 Hiarcs 5.0 Pentium 90 MHz               2428   38   363   59%  2363
>  32 MChess Pro 6.0 Pentium 90 MHz           2414   17  1669   46%  2441
>  33 Genius 3.0 Pentium 90 MHz               2409   26   785   61%  2329
>  34 Rebel 6.0 Pentium 90 MHz                2408   31   540   60%  2336
>
>
>
>
>
>>
>>Tiger 11.8 P90 -> 2483 (+54,-52)
>>Genius5 DOS P90 -> 2431 (+18,-18)
>>
>>(number between paranthesis are error margins)
>>
>>That means that with a good confidence:
>>
>>2431 <= Tiger 11.8 P90 rating <= 2537
>>2413 <= Genius5 DOS P90 rating <= 2449
>>
>>In the worse case Tiger 11.8 is 18 points below Genius5, in the best case it is
>>124 points above.
>>
>>The most probable is that Tiger 11.8 is 52 points above Genius5.
>>
>>I expect Tiger 12.0 to be at least 25 points above 11.8, so you are probably
>>right: Tiger 12.0 is probably 75 elo points above Genius5.
>>
>>
>>
>>
>>>4)I know that top programs do many stupid things so I believe that it is
>>>possible to improve them significantly.
>>
>>
>>You are right.
>>
>>
>>
>>
>>>I told Amir Ban about the problems that I found and I believe that he is going
>>>to improve Junior significantly.
>>>
>>>It is more easy to see problems then to fix them but I believe that it is
>>>possible to improve the software by more than 100 elo and maybe even more than
>>>200 elo in the next 2-3 years.
>>>
>>>I know that Rebel-tiger programmers also expect at least 100-200 elo software
>>>improvement from the fact that they work together.
>>
>>
>>Yes. I think we can achieve this just by exchanging ideas and selecting the best
>>ones. I'm not even talking about finding new ideas, and we DO continue to find
>>new ideas. Exchanging ideas is a kind of brainstorming, and new ideas appear
>>from time to time.
>>
>>
>>
>>
>>    Christophe
Yes, I believe in new ideas in math and science.  (Math in software and science
in hardware)
In terms of science, huge leaps in the speed of the computer can happen.  Cable
News Network says the computer hardware could use light (the speed of light that
is).
I am delighted to hear you have found mistakes in the "latest greatest"
softwares.  Math is a universal language.  Chess I was told is the most popular
strategy game on Earth.
The "Light Computer" can make more manipulations (and obviously more
calculations) of the math.
Jonathan (8th message)



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.