bayes

[Brian Fistler]

On Fri, 2004-03-26 at 16:17, Brett wrote:
> ok i am running 2 MS systems side by side when i send an email from
> yahoo or aol one goes through fine with all spam checks the other
> goes through the spam checks and bayes nails it
>
> SpamCheck: spam, SpamAssassin (score=5.4,
>         required 5, BAYES_99 5.40)
>
> now the same email from the other system got listed
> from bayes with 2.25 now they are different versions
> but would there be that much of a difference between the scores
> and is there any thoughts on controlling it\

Bayes is a LEARNING filter, and unless you trained both baysian filters
with exactly the same messages, and do not have SpamAssassin's
auto-learning turned on, each of the two systems will interpret the test
message you sent differently.

The larger the variety of messages that you manually submit to the
filters in training, the less variation you are likely to see between
the two.

If you look closely at the two systems you should see that the system
that allowed the message through likely did NOT give a BAYES_99 but
rather a BAYES_70 reply.

Send some spam and ham through the filter to train it, it seems like you
probably need more ham reports.

Brian