bug-gnubg
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bug-gnubg] Question: Interpreting Evaluation Results


From: Christian Anthon
Subject: Re: [Bug-gnubg] Question: Interpreting Evaluation Results
Date: Mon, 2 Mar 2009 22:29:04 +0100

If you are analysing an entire match or game it is normally not
necessary to go beyond the world class settings. The higher plies are
only marginally better, if at all, and consumes too much time. After
that you may/should use 4ply or rollouts on the interesting decisions.

Christian.

On Mon, Mar 2, 2009 at 9:29 PM, Bob Hart <address@hidden> wrote:
> Christian-
>
>        Thanks. My goal is to have the software consistently show the best
> move of the top three choices (or so) at the top of the list. This is so I
> don't have to look at thousands of results and carefully read the entire
> list for each position, thinking about plies and sort order. I am trying to
> understand how to do this before I spend a significant amount of time
> setting up multiple machines to have gnubg evaluate a large number of
> positions.
>
>        So, can I force the 5-ply evaluation to consider more than one move
> using the "Move Limit" setting in Settings/Analysis? Or should I do this
> with an even number of plies to get around the odd-ply/even-ply problem you
> mention? Or . . . ?
>
>        Thanks for any insight.
>
> -Bob
>
> -----Original Message-----
> From: Christian Anthon [mailto:address@hidden
> Sent: Monday, March 02, 2009 3:16 PM
> To: Bob Hart
> Cc: address@hidden
> Subject: Re: [Bug-gnubg] Question: Interpreting Evaluation Results
>
> It is generally not advisable to compare evaluations using different
> parameters. This is particularly true for gnubg which has an
> odd-ply/even-ply problem. In short, you should evaluate all the moves
> you wish to compare at the same ply.
>
> Christian.
>
> On Mon, Mar 2, 2009 at 8:29 PM, Bob Hart <address@hidden> wrote:
>> Hi-
>>
>>  I would like to understand how to interpret evaluation results presented
>> with different plies and out-of-order equities.
>>
>>  In reviewing the results below, I would ideally want to conclude "at a
>> glance" that gnubg evaluates 24/21 13/9 as the best play, because it is at
>> the top of the list. It appears, however, that the moves are not sorted
>> primarily in equity value, but rather by ply-value and then by equity
> value.
>> So is gnubg saying that the best move is the first one, 24/21 13/9, or is
>> the best move the third one, 24/20 13/10, which has the best equity?
>>
>> 1 Cubeful 5-ply -0.199          24/21 13/9
>> 2 Cubeful 4-ply -0.163  +0.036  13/10 13/9
>> 3 Cubeful 2-ply -0.155  +0.044  24/20 13/10
>> 4 Cubeful 2-ply -0.182  +0.017  24/21 24/20
>> 5 Cubeful 2-ply -0.217  -0.018  13/9  8/5
>> . . . Many more moves, all with worse equity . . .
>>
>>  Thanks for any insight.
>>
>> -Bob
>>
>>
>>
>>
>>
>>
>>
>> _______________________________________________
>> Bug-gnubg mailing list
>> address@hidden
>> http://lists.gnu.org/mailman/listinfo/bug-gnubg
>>
>
>
>
> _______________________________________________
> Bug-gnubg mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/bug-gnubg
>




reply via email to

[Prev in Thread] Current Thread [Next in Thread]