Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: SSDF DF7 K6 vs DSj 1.5a A1200 (22 - 24) Conclusion?

Author: Ulrich Tuerke

Date: 14:35:17 01/01/04

Go up one level in this thread


On January 01, 2004 at 13:54:00, Kurt Utzinger wrote:

>On January 01, 2004 at 12:50:56, Ulrich Tuerke wrote:
>
>>On January 01, 2004 at 09:24:39, Kurt Utzinger wrote:
>>
>>>Is this
>>>
>>>a)
>>>A further example that hardware is less important
>>>than most people think
>>>
>>>or
>>>
>>>b)
>>>a sign that Deep Fritz 7 is very much stronger
>>>than Deep Sjeng 1.5
>>>
>>>or
>>>
>>>c)
>>>other reasons?
>>
>>My conclusion is that either program is able to beat the other one.
>>
>>Any ELO difference derived from this match alone will vanish in the statistical
>>uncertainty.
>>
>>Uli
>
>
>     Hi Uli
>     True in principle but have you considered the difference
>     in hardware used?

Yes, I had realized it. I had been referring the hardware-software combinations.
Extrapolating this result to equal hardware results in a probabiltity that Fritz
may be stronger.
However, one should be aware that in view of the number of games (46) the
statistical uncertainty is still larger than 100 ELO.

A numerical example. Let's say the result on the above hw was 50%. Then a rough
estimate is that Fritz has a 70 ELO advantage (resulting from the hw
disadvantage). With a statistical error larger than 100, one can still not
exclude that Sjeng is stronger - even on same hardware.

However, the result seems rather to indicate some advantage for Fritz.
Nevertheless, I think that Sjeng's result in this match isn't bad at all. 50 -
80 ELO behind Fritz is a good result for a newcomer, imho.

Uli



>     Kut



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.