Author: Christophe Theron
Date: 01:12:06 12/17/00
Go up one level in this thread
On December 16, 2000 at 19:19:00, Eugene Nalimov wrote:
>You are really saying that today it's possible to write a program for a 20-years
>old computer that will be 40*20==800 Elo points stronger than a 20-years old
>program? I.e. 800 points stronger than Sargon or Rebel for Aplle II *running on
>the same hardware*?
>
>Eugene
Several points here:
1) Walter was talking about the software improvements in the last years, not in
the 20 last years. I think software improvements have been much more important
in the nineties than in the eighties for example (but maybe other people will
not agree).
2) The software improvements also take into account taking advantage of the new
hardware. There is no point in using hash tables on an Apple ][, so this
software improvement would not help on this computer. Ditto for endgame
tablebases. Hey, you know that very well, don't you? ;)
I guess null move is also pointless on a very slow computer. When you can only
reach 4 to 5 plies at tournament time controls, I doubt null move would help
much (and would surely introduce a lot of tactical flaws, as Bob repeats often).
I don't think current programs would be 800 points above Sargon on Apple ][, but
they would be dramatically stronger (on an Apple ][).
But port Sargon II on an Athlon 1GHz and let it play against current top
programs, and you will see a tremendous difference. Maybe not 800 elo points. It
is probably more like 500 or 600 elo points.
I think it is possible to make the experiment with Sargon III for PC, which was
stronger than Sargon II, but not dramatically stronger.
Christophe
>On December 16, 2000 at 17:53:24, Christophe Theron wrote:
>
>>On December 15, 2000 at 08:51:53, walter irvin wrote:
>>
>>>On December 15, 2000 at 05:59:54, Severi Salminen wrote:
>>>
>>>>On December 14, 2000 at 21:55:24, walter irvin wrote:
>>>>
>>>>>where will the next elo jump come from hardware or software??????????
>>>>
>>>>What do you mean? What was the last elo jump you saw? What is an elo jump? 1
>>>>elo, 10 elos 1000 elos? The answer to your question must be: from forest in form
>>>>of the Elo Rabbit!
>>>>
>>>>Severi
>>>your response proves one point that i have believed for a while .that there
>>>really has not been any measureable improvements by chess programmers lately
>>>.the old saying that if you keep doing what you are doing you will keep getting
>>>what you are getting applies to chess programming .i dont think that is going to
>>>change so i believe hardware is the only chance for a elo jump which i would
>>>define as at least 150 elo .
>>
>>
>>
>>When is the last time you have seen a 150 elo jump?
>>
>>Each year programs improve from 30 to 50 elo or more (of course not taking
>>hardware improvements into account) and you dare to say that "there really has
>>not been any measureable improvements by chess programmers lately"???
>>
>>Thanks a lot for supporting our efforts.
>>
>>
>>
>> Christophe
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.