> Whatever I do, sa-learn will not accept it as spam. :(
It takes about 500-1000 learned spams before SA starts marking things with the bayes filter. I've put at least that many in, and still only about half of my spams have bayes markings in them (and about 10% of my daily spam still gets through).
* Are you learning a proportionate amount of ham as well ? You can't just learn spam and not ham and expect bayes to work well, as it works by comparing the frequency of words in ham vs spam.
* Bayes doesn't "learn" particular messages, it learns the statistics of the words used in the messages. Eg how often the word "FREE!!" appears in spam, vs how often it appears in ham. If a spam doesn't have distinctive words that aren't used much in ham, bayes may have trouble identifying such messages even after training on them.
* If you're getting 10% of your spam getting through, you need to check other things as well as bayes too. Whats you're default threshold ? Are you using RBL checks ? Razor ? DCC ? Pyzor ? Bayes autolearning ?
Regards, Simon
------------------------------------------------------- This SF.Net email sponsored by: Parasoft Error proof Web apps, automate testing & more. Download & eval WebKing and get a free book. www.parasoft.com/bulletproofapps1 _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
