On Wednesday 02 June 2010 08:59:12 Olivier Teytaud wrote:
> > Any comment on the corresponding phenomenon in chess ?
>
> Sorry; I wanted to write "with alpha beta" and not "in chess".
Good question.
There has been quite some research on what happens as the programs go deeper
("DarkThought Goes Deep" and the successors such as "Crafty Goes
Deep", "New results in deep-search behaviour", etc.) and I believe the
consensus is that there is a diminshing return to searching an extra ply
deeper.
However, if you get your effective branching factor below 2, you will search
more than 1 ply deeper per doubling of speed, so you start negating the
above again. The effective branching factor of many modern programs is
around 2.
So the only correct way would be to observe the actual strength vs time (or
speed) evolution. I don't know of any in-depth studies on that.
--
GCP
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
