Most autolearn-only bayes databases wind up being mostly poisoned and wind up doing more harm than good. If you're seeing any spam with BAYES_ scores under 20, or ham with bayes scores over 80 you've got big bayes database problems. Spam under 50 or ham over 50 should also be extraordinarily rare. Most should wind up on the right side with either 00 or 10 for ham, and 99 or 90 for spam.
Replying to myself, My characterization of "no spam under BAYES_20" is a bit extreme, but not entirely off base.
For reference, here's my current BAYES_ distribution for all the tagged and false-negative spam I've got on hand.
1106 - BAYES_99 94 - BAYES_90 25 - BAYES_80 28 - BAYES_70 25 - BAYES_50 79 - BAYES_50 7 - BAYES_40 8 - BAYES_30 8 - BAYES_20 4 - BAYES_10 7 - BAYES_01 10 - BAYES_00
46 - no BAYES_ match at all.
So of 1,447 spams, 21 (1.45%) of them had bayes scores in the 00, 01 and 10 range.
------------------------------------------------------- This SF.net email is sponsored by: IBM Linux Tutorials. Become an expert in LINUX or just sharpen your skills. Sign up for IBM's Free Linux Tutorials. Learn everything from the bash shell to sys admin. Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
