My typical antispam arsenal consists of SpamAssassin plus Eudora’s Bayesian filtering. On a normal day in the past, SpamAssassin (set to a threshold of 6.5) would return around twenty false negatives. Of those twenty, Eudora might catch 75% on a good day. The rest would be marked as spam in hopes of further training the Bayesian filter.
Since installing these new custom rulesets, however, only two spams have gotten past SpamAssassin. Oddly enough, Eudora didn’t catch them either, but that’s beside the point. Any reduction in spam volume means more time for doing other things, so I’m very pleased.