This is just another example of why spam filtering methods should be combined. My bogofilter has a nice database of spam phrases, but it would have marked more spam as "maybe spam" if my email provider didn't also tag all email with spamassassin. It also helps to combine this with several RBL services, and maybe some hand-made identification regexps in procmail (or maildrop).
Using just one method at a time is no longer enough. It was a good thing when spamassassin introduced bayesian filtering, but by default it assigns way too low scores to be really useful.
Slashdot Mirror
Have a look at http://nexgenwars.com/, the numbers on that site appear to be pretty believable.
This is just another example of why spam filtering methods should be combined. My bogofilter has a nice database of spam phrases, but it would have marked more spam as "maybe spam" if my email provider didn't also tag all email with spamassassin. It also helps to combine this with several RBL services, and maybe some hand-made identification regexps in procmail (or maildrop).
Using just one method at a time is no longer enough. It was a good thing when spamassassin introduced bayesian filtering, but by default it assigns way too low scores to be really useful.