[messages] [General Discussion] Modulo bias

Joel Uckelman uckelman at nomic.net
Tue Oct 20 15:15:50 CEST 2015


Thus spake Filip:
> 
> "VASSAL 3.2.15 released" wrote:
> > * 11404: Some uses of the PRNG introduce a minuscule amount of modulo
> > bias
> 
> 
> Which uses of the PRNG in particular?

Any roll of a die having a number of faces which is not a power of two.

>  Was this issue 3.2 specific? Perhaps Windows specific or something?

No, the issue existed in all versions of VASSAL to that point, and
was not platform-specific.
 
> Some of my players used to be suspicious about the generator in an
> earlier version of Vassal and I did some testing. However, repeatedly
> rolling large numbers of dice (randomized layers) and checking reports
> proved that the distribution of results was ok. Or so it seemed. Since
> my private fork is still based on Vassal 3.1, I'm curious which uses of
> the PRNG actually produce this problem.

To give you a concrete example, if you rolled about 715 million six-
sided dice, the bug we corrected would give you on average one fewer
each of 5 and 6 than you should expect. It's exceedingly unlikely that
any single individual has rolled enough dice with VASSAL that this
would have affected them.

I fixed this bug because it annoyed me that we were doing something
incorrect, not because it could have caused anyone a problem.

-- 
J.


More information about the messages mailing list