Hi,
I have been receiving a lot of questions about MM recently, so I have
just updated my web page:
http://remi.coulom.free.fr/Amsterdam2007/
You'll find the ICGA Journal version of my paper there, with more
details about the math. You'll also find the C++ source code of my
implementation
David Fotland wrote:
Well I have no idea how much I gained from this. It might be weaker than
what everyone else is doing, since it seems I didn't implement this as it's
been described recently. My progressive widening only uses Rave values.
It's very simple. Others seem to have much more
David Fotland wrote:
Well I have no idea how much I gained from this. It might be weaker than
what everyone else is doing, since it seems I didn't implement this as it's
been described recently. My progressive widening only uses Rave values.
It's very simple. Others seem to have much more
So far on 9x9 go, Many Faces doesn't seem to make a huge difference. On
19x19 it makes a huge difference. I ran test games overnight against Gnugo.
With Many Faces turned on, my engine wins 85%. With many Faces turned off,
my engine wins 7%.
Both results are unexpected. Since most of my
I use rave to decide which moves to promote into the pool of moves being
evaluated by the UCT formula. I use Many Faces to bias the rave values.
I haven't tried other approaches so I can't say how or why my approach works
well. Certainly it doesn't scale very well to huge numbers of playouts,
What's your general approach? My understanding from your previous posts is
that it's something like:
UCT search using Silver's beta formula and UCB1 with Win-rate and Rave for
choosing a child (I use basic UCT with win-rate and Rave, and the original
MOGO beta formula).
UCT search is biased
I was not at all surprised by this result.
My thinking goes like this. On 9x9 the global situation is everything
that matters and precomputed information is not as important as
searching effectly is. Good 9x9 games are often very sharp fights
where then next move often violates good shapes
For 9x9 games, when I added progressive widening to AntiGo (before I added
RAVE), it was low hanging fruit. I used my old program Antbot9x9 for the move
ranking and got a very nice strength increase for very little effort. Then,
with a bit of tweaking, I got more improvement. RAVE, on the
I made some more tables with 50K playouts, in case that is easier to look at.
50 K playouts?
LIGHT PLAYOUTs XX
rave
?? 1|?? 36?? 39?? 38?? 40?? 39?? 40?? 40?? 40?? 37
?? 2|?? 38?? 40?? 42?? 42?? 42?? 42?? 42?? 40?? 40
?? 3|?? 40?? 42?? 42?? 43?? 43??
