Author: Chris Carson
Date: 11:07:19 07/07/00
Go up one level in this thread
On July 07, 2000 at 02:26:12, blass uri wrote:
>see
>http://www.kasparovchess.com/serve/templates/dispchessdoc.asp?p_docID=7941&p_docLang=EN.
>
>They are sure that Deep Junior has at least 2640.
>They claim that this is the performance of Fritz in the dutch championship.
>
>I think that they are wrong because I read that Fritz failed by 1/2 point to
>achieve a GM norm and performance of 2640 is enough for a GM norm.
>
>They also say that Deep Junior has more than 2900 by the ssdf if you consider
>the speed advantage.
>They seem not to know about diminishing return.
>
>I do not believe that Deep Junior with 8 processors can get 2900 ssdf rating
>because of diminishing returns.
>
>Diminishing returns was proved in fritz-fritz games and are probably right also
>for Junior.
>
>Uri
Good post Uri. Here is my take on the event. Note: I have not talked with any
of the club kasparov people, so all this is pure speculation on my part. :)
Look at my list: http://home.interact.se/~w100107/may.htm
Fritz multi has a rating of 2635 plus or minus 100 points. The TPR from the
Dutch Championship was 2595 (with 2 games removed due to forfeit) and one game
frone game from the last WCCC included. So 2640 is not a bad estimate for Fritz
on a 4x500mhz machine and neither is 2540 or 2740 since both represent one error
of measure from the estimate. :) This is close to the Club Kasparov number.
The Fritz rating could even be higher for one or both of the following reasons:
1. All games counted including forfeits.
2. At least one draw by Fritz might have been a win on time if the operator
had not agreed to a draw
However, I think the 2635 is a fair reflection of Fritz rating and as I recall,
the NPS were about twice the NPS on a 500mhz machine (due to overhead of
multiprocessing).
So what does this say about Junior. Perhaps nothing.
Deep Junior has a multi processor rating of 2519 on dual processor systems. I
think on a 8 processor machine you can expect 75 to 100 point gain due to the
higher number of processors (I am not sure of this for Junior, since I have no
hard data, just a gut feel). So, 2600 plus or minus 100 points would be in my
opinion a reasonable guess at the TPR for Deep Junior. I expect a performance
between 2500 and 2700. My guess is that Junior will win 2 games on time
(perhaps 3) and may pull one upset win or draw. I wish Deep Junior luck and the
GM's the best. This is truly a great event! :)
BTW. Chess Life (July 2000) has a great write up on the Dutch Championship in
the Letters to the Editor (Europe) section.
I plan to publish an updated Ratings list and a Summary TPR for all the
tournaments played so far after this is over. :)
Best Regards,
Chris Carson
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.