Brian Nelson wrote: > The scoring in more recent versions of spamassassin is generated using a > genetic algorithm that finds the optimal success rate while keeping > false positives to a minimum. It's not something I'd mess with lightly, > unless you really know what you're doing.
How does the generic algorithm know the positive/negitive rates of the
Bayesian scoring at my site? Granted I'd not touch the other scores but many
people, myself included, have found that the Bayesian score is often accurate
enough to go with alone. That is not reflected in SA's scores, at least not
up to the installation I have, where all but the 99-100% range has a high
enough score to barely nudge a message into the spam range. Anything less
than that and in the absence of all other spam markers (which is common on
some spam as they test against SA) the message would get through as a false
positive.
--
Steve C. Lamb | I'm your priest, I'm your shrink, I'm your
PGP Key: 8B6E99C5 | main connection to the switchboard of souls.
-------------------------------+---------------------------------------------
signature.asc
Description: OpenPGP digital signature
