dear Go researchers,

>> > Found a 582 move 3x3 game...
>> Can you give us sgf?
>
> I took the effort of trying to format the 582 game in a more insightful way.
> I ended up with lines of positions that mostly add stones, only starting
> a new line when a capture of more than 1 stone left at most 4 stones:
> The result is attached. I think there is clearly
> room for improvement, i.e. make this game much longer.

Gunnar Farnebäck took up my challenge and took a giant leap from 582
to 1808 moves, using a UCT oriented search with maximum playout length
as the score.

That led me to implement a so-called beam search,
in which we keep a set of W (the beam width) games at the same depth
D, and for each one, play some number B (branching factor) of random
playouts.
We then keep the W longest of these W*B playouts, truncated at depth D+1,
and repeat the process, until no more playouts are possible (all moves
being superko violations).

I ran a dozen beam searches with W=16384 and B=1024, and the best one
reached a record 2900 moves. See the attached sgf.

It's quite mesmerizing to load this sgf in your favorite go editor and just
cycle through all the moves at a rapid pace...

The rules should really be TT (Tromp/Taylor) instead of NZ (New Zealand),
but unfortunately Cgoban doesn't support TT.

regards,
-John

Attachment: 2900.sgf
Description: application/go-sgf

_______________________________________________
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Reply via email to