Re: [computer-go] A superko paradox

[Chris Fant <chrisfant@xxxxxxxxx>]

> I'd say that a 128 bit Zobrist hashing with random values has
> entirely negligible odds of generating a false collision;
> At 64 bits the odds are sizable, but probably negligible compared
> to some other concerns, like DRAM errors on a PC without ECC,
> or plain software bugs.

Plus, a collision is not all that bad.  Not making the move that leads
to what is thought to be a repeated board is only a bad thing when the
move you wanted to make was correct and urgent.  Obviously that will
happen a lot more often if the check is performed during search.
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/