[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Bug-gnubg] A new variance reduction technique
From: |
Christian Anthon |
Subject: |
Re: [Bug-gnubg] A new variance reduction technique |
Date: |
Tue, 31 Mar 2009 12:33:52 +0200 |
Without consulting the code I believe that luck-lookahed is used in
three places. Analysis (configurable), best-roll-dice (not
configurable, 0ply, cubeful?), rollout variance (not configurable,
0ply, used to be cubeless, now cubeful).
Christian.
2009/3/30 Massimiliano Maini <address@hidden>:
>
> Michael Petch <address@hidden> wrote on 30/03/2009 16:30:24:
>
>>
>> On 30/03/09 8:14 AM, "Massimiliano Maini" <address@hidden
>> > wrote:
>
>> 2. I'm not sure gnubg uses 2ply for luck evaluation. If it uses 0
>> ply (as I remember,
>> but I could be wrong, mabe it is 0ply just for match analysis luck
>> evals), introducing
>> your trick would cost something in terms of speed (actually, quite a
>> lot I would say,
>> knowing that each additional ply cost more or less a factor 21). If
>> somebody on gnubg
>> mailing list can confirm that ....
>>
>>
>> By default luck is analyzed at 0 ply. You can set the plies manually with:
>> set analysis luckanalysis plies <plies>
>
> Right, but is this setting acting only on the luck analysis (the luck
> adjusted
> result you see in the stats of an analyzed metch/session) or also on the
> variance
> reduction algorithm ?
>
> MaX.
> _______________________________________________
> Bug-gnubg mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/bug-gnubg
>
>