If white plays first from this position, Valkyria plays J2 which wins
100% of time.
If black plays first, black has initially sometimes 5% chance to win
for the best move but the chance goes down to 1% rather quickly.
I admit not reading this thread carefully enough to understand what
the argument is about. So I just contribute with following statement:
Positions like this are very easy to handle for MCTS based programs
that uses heavy playouts.
Ok, some other thoughts:
Valkyria need only 375 playouts/per move to match the playing strength
of GnuGo on 9x9. I have been doing some scaling studies with Valkyria
recently I find interesting. The unresolved question for me is if
playouts should be really simple and fast, or if one can gain from
slower heavy playouts. I will follow up some with my data but now I
need to go to bed!
-Magnus
Quoting terry mcintyre <[EMAIL PROTECTED]>:
From: Gunnar Farnebäck <[EMAIL PROTECTED]>
Darren Cook wrote:
> What do your program's playouts think when presented with the board
> position in the article? This is a terminal position, both players have
> passed, a comfortable white win, yet pure random playouts think black
> will win more often.
>
>>> http://dcook.org/compgo/article_the_problem_with_random_playouts.html
Any chance to get an sgf of the position?
Are you referring to this one: "A position of the third type is
shown to the right".
(;FF[4]CA[UTF-8]AP[GoGui:1.1.1]SZ[9]
KM[7.5]DT[2008-10-15]
AB[hb][ib][ic][id][ie][he][hd][gd][fc][ec][dd][de][ee][fe][ff][gf][ig][eg][dg][cg][cf][bf][be][ae][ad][af][ch][eh][fh][di]
AW[ia][ha][ga][gb][fb][eb][db][bb][ac][bc][dc][hc][gc][bd][cd][ce][ef][fg][gg][hg][hf][if][hh][gh][gi][fi][ei][ag][bh]
PL[W];W[];B[])
Does this list allow attachments? I know many lists are filtered. Is
it configurable to permit .sgf attachments? Unlikely to carry
viruses.
_______________________________________________
computer-go mailing list
[email protected]
http://www.computer-go.org/mailman/listinfo/computer-go/
