[Computer-go] monte carlo search; all valid moves?

Urban Hafner contact at urbanhafner.com
Sat Mar 28 03:41:00 PDT 2015

20 playouts is a bit slow Folkert. But I've been there, too! Now my bot is
at around 1000 19x19 pps and 5000 on 9x9 (but I also check to not play into
single point eyes). That's still quite slow and further optimisation will
be required. But OTOH on a 4 core machine without UCT (only the AMAF)
heuristic I get around 1250ELO on CGOS 13x13 (


On Sat, Mar 28, 2015 at 11:30 AM, Petr Baudis <pasky at ucw.cz> wrote:

> On Sat, Mar 28, 2015 at 10:15:32AM +0100, Álvaro Begué wrote:
> > I am not sure I understand the question. The only thing that is typically
> > not checked in the playouts is superko. What other "validity checks" are
> > you performing?
> There is a few:
> (i) No single stone suicide.  Can't think of how to playout without
> that, you wouldn't know when to stop playing.
> (ii) No multi stone suicide.  Sometimes in the fastest implementations
> this rule is relaxed (i.e. you proclaim that you are using Tromp-Taylor
> or New Zealand rules).  But in real-world engines, you need to check
> things like self-atari or 2-liberty semeai and then a suicide check
> naturally falls out from that too.
> (iii) No ko. This is a pretty cheap check, and without it again your
> playouts may end up looping.
> BTW, the fastest playouts are probably done by libego and that's open
> source.  But as I said in the past - consider not optimizing first;
> write some reasonable playout heuristics first and *then* profile your
> code.
>                                 Petr Baudis
> _______________________________________________
> Computer-go mailing list
> Computer-go at computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go

Blog: http://bettong.net/
Twitter: https://twitter.com/ujh
Homepage: http://www.urbanhafner.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://computer-go.org/pipermail/computer-go/attachments/20150328/ae53c7e6/attachment.html>

More information about the Computer-go mailing list