[computer-go] 19x19 Go, scalability with time vs handicap

[Sylvain Gelly]

Hello Daniel,

With the addition of fuseki and joseki library will its rating increase?
> Especially a fuseki library. Will a fuseki library be consistent with its
> playing style?
>
That is an interesting and not trivial question. The problem is that the
player has somewhere to understand the opening not to destroy it after. The
problem is not that MoGo is playing some strange moves, it is that it is
happy with them :-).

For human players a difference of 2 kyu means that the winning ratio of the
> stronger player is almost 100%.
>
Is it? Do you have some statistics? If so, that is interesting, because that
means that neither MoGo nor GnuGo exploit well (comparing to humans) the
handicap stones (see results of handicaps with settings which make them even
at H0).


> My feeling is that it's going to take more than double of speed for MoGo
> to reach 3kyu.
>
Actually my goal was not to argue about how to reach some rating, but what
happens when giving more time against a fixed player with handicap. Also, I
feel like the KGS rating is not well adapted to bots rating, because of the
inertia. Bots are playing so many games a week that once they reach one
stable rating, the rating will change very slowly.
Also, strange results happen in MoGo vs human games: white almost always
wins! That means the stronger player, no matter the handicap, manages to win
the game. That is quite asymetrical. BTW it was one of the reason I launch
those experiments, expecting asymetrical results when playing black or white
with handicap. This did not appears in those experiments.

Sylvain
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://computer-go.org/pipermail/computer-go/attachments/20070422/894db83b/attachment.htm