[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: R?f. : [Bug-gnubg] Randomizing GNU-bg's first move.
From: |
Jim Segrave |
Subject: |
Re: R?f. : [Bug-gnubg] Randomizing GNU-bg's first move. |
Date: |
Thu, 28 Aug 2003 16:20:17 +0200 |
User-agent: |
Mutt/1.4i |
On Thu 28 Aug 2003 (13:09 +0200), address@hidden wrote:
>
> >One quirk about improving your backgammon by playing against bots is that it
> >will always play the same opening move at a given setting.
> >
> >E.g. for money att 2-ply, GNU-bg will always play 21 opening 13/11 24/23 etc
> >etc.
> >
> >The alternative of 13/11 6/5 is likely equally as strong but leads to
> >different types of games.
> >
> >All this leads me to wonder if an option to randomize GNU-bg's opening move
> >might be an interesting option.
> >
> >My ideas was to select a threshold EMG or MWC within which GNU-bg could
> >randomly select a different opening play. To take it further one could select
> more
> >than just the opening play.
> >
> >Anyone else apart from me find this useful?
>
> I do (do do, da da da).
> As a minor detail (maybe annoying to implement), analyzing GNU when his 1st
> move has been "randomized" should take the fact into account (maybe do the
> same reasoning and not counting the eventual equity difference from top
> ranked move if the difference is below ther EMG/MWC threshold of the "1st
> move randomizer").
My inclination would be to consider this a pretty obscure feature
which is already do-able.
Set player 0 to human, start match, pick what will be gnubg's first
move by hand. Then set player 0 back to gnubg. You've now achieved the
first part without adding noise or anything else.
--
Jim Segrave address@hidden