Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: CEGT: testing and presentation of results

Author: Uri Blass

Date: 23:10:58 10/19/05

Go up one level in this thread


On October 20, 2005 at 02:02:18, Heinz van Kempen wrote:

>On October 20, 2005 at 01:47:42, billiau wrote:
>
>>Hi,
>>
>>I am agree about 99% of your post (hard work, errors, etc...).
>>
>>About the 1% chance for a setting to be better, if they were so sure,
>>i can't understand why they let you test the Fruit setting.
>>
>>Thurthermore, if the setting is only approximatly equal to the standard one in
>>long time control and it is better in analyzing mode (test suite...), i think it
>>could be better to officialize the setting.
>>
>>G. Billiau
>
>Hi,
>
>what you tell about analyzing mode is of course a point. The question is if Uri
>also tested positions where default setting found superb moves and the setting
>not. I asked this before.

There are of course positions when the default setting is faster in finding
moves but the feeling based on test suites is that the general result is better
for selectivity=50

I also remember that in one game that Fruit Uri lost Fruit default needed more
time to avoid the losing mistake.

In most games that I looked at them I did not detect the losing mistake so based
on my memory so I tested only 2 positions from the CEGT games when in one of
them I could not reproduce the error of Fruit Uri.

Uri



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.