gnugo-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [gnugo-devel] Dragon safety early in the game


From: Arend Bayer
Subject: Re: [gnugo-devel] Dragon safety early in the game
Date: Wed, 28 Aug 2002 17:33:55 +0200 (CEST)

On Wed, 28 Aug 2002 address@hidden wrote:

> GNU Go often neglects groups in the opening or early
> middle game. Examples of what I mean are century2002:50,
> or the tenukis in games/cgf2002/gnu-har.sgf at moves 19 and 21.
> GoFuN was often able to surround a group at a distance then
> kill it. And this happens all the time in 9 stone games
> against strong players.
>
> If we can identify these situations then we can increase
> the strategic effect of defensive moves and fix this problem.
>
> One possibility is that we try to use the dragon2.weakness field,
> which Arend introduced around 3.3.4 to identify these weak dragons.
> This field needs to be better documented I think. As far as I know
> it isn't used much yet and maybe Arend has ideas about using it
> more. We should figure out whether that is an adequate measure
> for this problem, and if not, maybe it can be improved.

Of course dragon2.weakness was introduced exactly to address this
problem. I still think that it is a useful way to go, but it definitely
needs further tuning. I know I have to add documentation, and it has to
go in the -w output, too.

I disagree about the fact it isn't used much. The strategic effect on
a dragon is now computed basically as

        0.75 * dragon.effective_size * dragon2.weakness.

The 0.75 probably doesn't make too much sense any more.

E.g. in century2002:50, the weakness of E4 gets estimated as 0.72 (the
weakness can range from 0.0 to 1.0). (This information is available with
dragon_data E4 from GTP.) This looks reasonable to me.

The strategic effect of F7 on E4 is then computed as
0.75 * 10.13 (effective size) * 0.72 = 5.49.

This does not look reasonable to me. Hence we have to tune the above
formula or the computation of the effective size. Another thing to note
that defending E4 here is mostly interesting because J4 isn't alive yet
(and it has a reasonable .weakness of 0.78). Maybe a strategic defense of
a weak dragon could automatically trigger a strategic attack of a
weak hostile neighbour.

Certainly there is a lot to do.

Arend






reply via email to

[Prev in Thread] Current Thread [Next in Thread]