Abigail Marshall <[EMAIL PROTECTED]> writes: > Also, you have said that SA parses out the Base64 comment, but the > post that started this discussion came from someone who had a heavily > porn-laden piece of spam come through, and their SA failed to register > any tests on the body.
A sample size of one is not very meaningful. We're running nightly tests on a corpus of 220,603 messages. We tuned the scores for SA 2.50x on a corpus of 258,304 messages. I think changing a score based on a missed false positive or two, even if it's on your own email is about the same thing as believing that you're more likely to win the next spin at the roulette wheel after a losing streak. Anyway, you need a larger sample size, of both ham and spam, before you can conclude that raising a score is going to improve your overall results. I've only upped one score in my user preferences (setting MIME_HTML_ONLY to 1.5), but only because it's vastly better for me than most people and I was able to test it over time on my email corpus (of 30,000 messages). Given the number of times that I've read that SA doesn't even decode base64, that we don't use Bayes, and so on, I find most anecdotal evidence given here very dubious. Daniel -- Daniel Quinlan anti-spam (SpamAssassin), Linux, and open http://www.pathname.com/~quinlan/ source consulting (looking for new work) ------------------------------------------------------- This SF.Net email is sponsored by: INetU Attention Web Developers & Consultants: Become An INetU Hosting Partner. Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission! INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk