RE: [computer-go] alpha-beta + graph history question

["Persson, Magnus" <persson@xxxxxxxxxxxxxxxxxx>]

I had a bug in my transposition code that is very easy to make and very
difficult to find until one suddenly sees the light.
 
When you finally store the result of a call to the alfabeta function in
the t.table, it is not certain anymore that the node in the table
anymore is the same as when alfabeta was initially called, because the
recursive alfabeta calls might have overwritten the entry in the table.

I could not tell from your code example if this could possibly be the
cause of your trouble but I think it is a kind of bug almost all
programmers would make unless you are lucky and have a t.table
implementation that keeps you from having this bug. If it does not apply
to you perhaps someone else suffers from it. If the table is large then
it might be a minor problem, but in my case it led to crashes and weird
inconsistencies that I found with assertions.

Best wishes
Magnus

--
Magnus Persson
Center for Adaptive Behavior and Cognition
Tel: +49-(30)-82406-350
Cell phone: +49 163 6639868

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