Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rybka 1.01 Beta 13 b First impressions

Author: Vasik Rajlich

Date: 14:58:37 02/02/06

Go up one level in this thread


On February 02, 2006 at 08:56:52, Albert Silver wrote:

>On February 01, 2006 at 17:49:03, Vasik Rajlich wrote:
>
>>On January 30, 2006 at 18:14:21, Albert Silver wrote:
>>
>>>>Let me put it like this: that is the chance at this point (ie. just based on the
>>>>games you list above, without any further testing) that your hypothesis is
>>>>correct?
>>>>
>>>>If you analyze it "statistically", you might get some figure like (let's say)
>>>>30%. Just a total wild guess, eyeballing your numbers.
>>>>
>>>>In reality, though, it's probably more like 3%.
>>>>
>>>>The reason is that before your experiment started, there were let's say 20
>>>>candidate hypotheses, that you didn't even bother to list. Maybe ultrasolid is
>>>>worse against Fritz, maybe it's better in closed positions, etc. One of these
>>>>hypetheses is likely to get lucky, and this hypothesis will then of course have
>>>>very nice data to support it.
>>>>
>>>>Anyway, there is nothing wrong with this procedure, as long as you eventually
>>>>test your hypothesis "straight up".
>>>>
>>>>Vas
>>>
>>>By "straight up", do you mean it is tested alone without any other parameters
>>>influencing? If so, wouldn't that go against the theory that each parameter is
>>>independent of the other and should bring its fruits?
>>>
>>
>>What I mean by straight up is that the testing should go in the following
>>sequence:
>>
>>1) Play a bunch of games, with various settings, without any special
>>expectations
>>2) Identify some trend - let's call it a "candidate hypothesis"
>>3) Test the candidate hypothesis with many many more games
>>
>>Maybe, to be really fair, the games from step 1 should even be thrown away, and
>>only the games from step 3 should be used. Not sure about this.
>>
>>Consider the following scenario.
>>
>>There is some person (Joe) who just won the lottery. You want to see if Joe is a
>>specialist at winning lotteries. So, Joe buys many more lottery tickets, and we
>>see how he performs. The question is: is it fair to include that first win in
>>his statistics, or not? Probably not.
>>
>>Vas
>
>Unless the testing conditions are identical (time control and openings), in
>which case there is no need to repeat the result. Otherwise, your description
>above is exactly what I had in mind, so we're seeing eye to eye here.
>
>                                          Albert

All right. Actually the initial exploration (ie. step 1) should ideally be done
with something besides games, but I don't know what. Maybe some sort of a "good
testsuite".

Vas



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.