[Computer-go] Criticality

Stefan Kaitschick stefan.kaitschick at hamburg.de
Tue May 28 10:35:54 PDT 2013


A while back Remi Coulom made a great proposal for measuring
"criticality", the importance of a local situation in contributing to
winning or losing.
Go programming has become less of a public affair, so maybe some
programs are utilizing it in some way.
But it's actually not easy to say what to do with the information in
the context of mc progamming.
Here is one idea: one of the most detrimental effects for bots is the
multiplication of group status errors.
If three too big to lose, but actually safe groups are evaluated at
80% alive each, the bot thinks the game is almost even, even though
its hopeless.
To combat this effect, three seperate local analyses could be made of
these three areas. A local analysis should be much more precise,
coming back with an almost perfect result. But what to do with the
information that a group "is not supposed to die" or "not supposed to
live"?
One simple use would be to give playouts that do not conform to the
postulated results a lower weight.
(I only mean conceptually simple, ofcourse)



More information about the Computer-go mailing list