Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Further testing of Genius 6.5: +0 =4 -46 against Ruffian

Author: Mike S.

Date: 22:54:20 10/26/02

Go up one level in this thread


On October 27, 2002 at 01:28:30, Uri Blass wrote:

>(...)
>>But OTOH, maybe the positional part of the WM-Test(100) indicates that the
>>extreme result isn't to unlikely: In the positional part of that test with 36
>>pos., Ruffian solved 22 - and Genius 6.5 solved only 7!!
>
>I do not think that we can use the WM-test suite to decide about playing
>strength of the engines. (...) If ruffian can score significantly better than
>genius in a positional test then soemthing is wrong in the test.

This would mean you know what a test result *should* be beforehand (which means
basically, you don't need to test at all :o).

I think we must not forget that our good impression, and the good reputation, of
Genius comes from a more or less *distant past*. It hasn't been used continuosly
much in matches against recent engines. Latest SSDF results I found:

Genius 6.5 128MB K6-2 450 MHz, 2473
Opponent Result
Fritz6 K6-450  10-30
Junior6 K6450  14½-27½
Shred5 K6-450  12-30

The WM-Test is known to be very difficult even by todays high standards. The
best positional result was 26 solutions of 36 by Fritz 7.0.0.8.

In general, I tend to doubt positional tests too... but I think it can't be that
wrong, 22 solutions vs. 7 solutions! If you take a look at the positional
results of >100 engines, it roughly follows the general engine's strength. If it
would be so wrong that 22 vs. 7 tells nothing, then the engine ranks in that
part would have to be chaotic - which they absolutely aren't. I can't claim that
these 36 positional tests are perfect, which is probably impossible and I also
didn't study them, but OTOH I'm sure that most of them are fairly ok. Much
effort has been put in the creation of that test.

http://www.computerschach.de/test/index.htm
(MS-Excel or Excel Viewer needed for the results.)

Regards,
M.Scheidl



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.