IIRC, Hendrik Baier's master thesis shows an experiment conducted using
Orego that establishes that Dynamic Komi improved Orego's results against
GnuGo. These are non-handicap games.

I recall that others have reported the same thing (that dynamic komi helps
even in non-handicap games). 

IMO, Nick's entropy argument provides good theoretical reasons why it might
work. The policy also has practical support, from the idea that adjusting
komi keeps the program playing aggressively.

Brian

-----Original Message-----

> In a Monte-Carlo program, the amount of information derived from one
playout is given by its entropy
>  -p(win).log_2(p(win)) - p(lose).log_2(p(lose)).
> This has a maximum at p(win) = 0.5, and is 0 if p(win) is 0 or 1.
> 
> Similarly, suppose its best move has won less than 10% of the playouts. It
could resign, but let's say it is giving a handicap to a weaker player.
Instead of just doing more playouts, it can pretend that it will be
receiving extra komi.  Again, the quality of the information per playout
then drops, but the quantity goes up, hopefully by more than enough to
compensate.
> 
> This seems like an argument for using dynamic komi, adjusted from time to
time during each game move.
> 
> Nick


_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go

Reply via email to