Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: What's Enriques test suite ELO prediction for J6?

Author: Enrique Irazoqui

Date: 08:23:56 02/02/00

Go up one level in this thread


On February 02, 2000 at 11:02:47, Howard Exner wrote:

>On February 02, 2000 at 06:42:52, Enrique Irazoqui wrote:
>
>>On February 01, 2000 at 22:13:43, Charles Unruh wrote:
>>
>>>What's Enriques test suite prediction for J6?
>>
>>T12 -51. As you see, it's way off.
>>
>>>What are the theories for the
>>>deviation between (ENRIQUE'S)test suite and ssdf?
>>
>>More than theories, what I have is a list of over 20 programs that perform
>>basically the same in my test and in the SSDF list. Junior 6 is the exception,
>>is the only program I checked whose strength is greater than its tactical
>>ability.
>
>If your test is mostly tactical

Exclusively tactical.

> then the contest engine for Rebel Century called
>Q3.eng would likely do much better than Rebel Century default on your suite. Yet
>playing Q3.eng vs Rebel Century in a series of games would most likely see Q3
>lose badly.

Sure. We already talked about this. It is possible to fool a tactical test suite
in 1001 ways. But programmers tune for real life, not for test sets, and that's
why the test works. You can argue and deny it, but it does work for all programs
so far, as posted several times*before* the SSDF list appeared. Not for Junior
6, though.

Enrique

> There seems to be a trade off. As an old favorite album title of
>mine by the Moody Blues says, it's "A Question of Balance".



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.