Re: [gnugo-devel] Move valuation question

[bump]

> Isn't this a bit exagerated ? I mean, let's imagine we're only 2.5 pts
> behind and a group worth 40 pts is in danger. Don't you think GG would be
> taking way too much risks in such case ?

Certainly, you are right.

> I'm not sure what to do, but what about following ?
> 
>         this_value = 2*dragon[aa].effective_size - gg_abs(score);
>         if (this_value < 0.0)
>           this_value = 0.0;

This doesn't look right to me, however. Suppose you are
fifteen points behind. Practically any dragon on the board
will have effective size 7.5 or more, so this statement
has no effect. Yet the right thing may be to gamble and
rip sente. Assume the uncertain result is right and go on,
or defend and fall further behind.

Of course sometimes a defensive move can be worth real
points. In that case, if we can detect this fact, we
should pay attention to the uncertain owl defense.

What we really need is an additional certainty
estimate---we need to know not only whether the owl
result is certain. We also need to know whether the
score estimate is certain, for example, how far we
are from the end of the game.

Dan