Re: [gnugo-devel] inconsistnies in owl_attack and owl_does_attack

[Evan Berggren Daniel]

On Tue, 13 Aug 2002, Evan Berggren Daniel wrote:

> Another option that doesn't fix anything but will probably improve play
> would be to run owl_does_attack on all the top moves proposed, and make
> sure that all attacks are being counted in the final valuation.  This way,
> if, say, the third ranked move has an attack that isn't found, with a high
> enough value to be worthwhile, gnugo will find it and play that move
> instead of the previously ranked top move.

It occurs to me that this is actually very similar to the proposal to just
play out the top couple moves in a 2-ply or larger search.  Essentially,
the problem is we don't search deep enough to see all necessary attacks,
and either approach would fix that problem.  If the resulting valuations
are correct in either method of doing things, which they should be, then
the only real difference between the proposals is breadth vs. depth.

To combine them, I would suggest that we do a broad search, ie look for
missing attacks on the top 5-10 moves, and then take the top 2-3 another
level deeper, check for missing attacks again, and then pick between them.

Thanks

Evan Daniel