On 05.10.2010 18:40, [email protected] wrote:
No information is thrown away with maximizing win rate.
:-)
This is because in go it is only the sign of the final board count
that matter.
It struck me as well reading this debate that it's all about using the
score as well as the end resut.
Dynamic komi tries to influence the agressiveness of play based on how
well you are doing, taking into account the likely higher (or lower)
skill of the opponent in a handicap game. Another way of doing this,
using more information, is to evaluate with both the real and the
komi-for-strength-adjusted result. For instance, if a move makes you
drastically *worse* off in the real win percentage, you don't play it,
regardless of its strenght with adjusted komi. Then I thought, why not
use more than one adjusted komi value? Then it struck me that this is
just doing statistics on the final score in the leaf nodes of the
playouts. The more komi values you consider, the less of the end score
you throw away.
Yes I know I am very stubborn and narrow minded on this issue.
Seems like most people here are, on most things. ;) But empirical resuls
win out in the end for most of us.
It is a weird thing to fear throwing away "information" and then
estimate the expected score which will throw away the actual
distributions of outcomes. And so will any other statistical measure do.
You don't have to throw away anything, keep both.
Also I do measure the score in at least two ways in Valkyria. One is
simply taking the average of score at the root and the other is to
estimate territory directly by looking at black/white membership of
individual points of the board. The latter is much more stable and the
average is often off several points also late in the endgame.
Have you considered (or tried) using the median, rather than the average?
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
