Actually, a better Go-God for handicap games would probably be one that
ignores score margin as long as it's behind and simply maximizes the
entropy measure for the lowest-entropy proof tree that proves that Black is
winning. (And only counts the entropy for the black moves, not the white
moves in that tree). Once ahead, of course it can do whatever it wants that
preserves the win.
On Sun, Aug 6, 2017 at 10:31 AM, David Wu <lightvec...@gmail.com> wrote:
> * A little but not really.
> * No, and as far as we can tell, never. Even 7x7 is not rigorously solved.
> * Unknown.
> * Against Go-God (plays move that maximizes score margin, breaking ties by
> some measure of the entropy needed to build the proof tree relative to a
> human-pro-level policy net), I guess upper bound at 5 stones, likely less.
> Against Go-Devil (plays moves that maximizes win chance against *you*, has
> omniscient knowledge of all your weaknesses and perfect ability to forecast
> when you would have a brainfart and blunder), unknown, probably a bit more
> than vs Go-God.
> On Sun, Aug 6, 2017 at 9:49 AM, Cai Gengyang <gengyang...@gmail.com>
>> Is Alphago brute force search?
>> Is it possible to solve Go for 19x19 ?
>> And what does perfect play in Go look like?
>> How far are current top pros from perfect play?
>> Computer-go mailing list
Computer-go mailing list