Re: [Bug-gnubg] Asking gnubg for the optimal move

[Michael Petch]

On 2013-10-22 12:54, Philip Jackson wrote:
> Hi all
>
> I'm starting an undergraduate dissertation in which I write a
> reinforcement learning algorithm which learns to play Backgammon. I
> need a decent opponent to train it against, and it's looking like
> gnubg is the best (and open source, which is great). Unfortunately I
> have no idea how to go about interfacing with gnubg - I need some way
> of sending it a board state and receiving an optimal move in reply (or
> a state evaluation). I'm pressed for time and the source looks pretty
> expansive, so I'm asking you for pointers - where should I start, is
> there an API that I can use, how difficult should it be, or is there
> another backgammon engine which would be easier to talk to?
>
>

There are a few approaches including the external socket interface;
python interface; or just launching GNUBG from a script/program and send
it commands via the command line and parse the output (gnubg-cli.exe for
Windows, and gnubg -t  for Linux/Unix). Some information I gave
previously is here:

http://lists.nongnu.org/archive/html/bug-gnubg/2013-06/msg00051.html

If you have any specific questions or problems just ask, but hopefully
something there will get you started.

-- 
Michael Petch
GNU Backgammon Maintainer / Developer
OpenPGP FingerPrint=D81C 6A0D 987E 7DA5 3219 6715 466A 2ACE 5CAE 3304