Jeremy Morton wrote:
... what do you guys get when you put this through your SA install?
http://rafb.net/p/0z7rXo65.html
Mine only gives me a Bayesian spam likelihood of 50%, but it seems to
use lots of spammy words so I'm not sure why it isn't a lot higher. Even
without training shouldn't it be higher?
Best regards,
Jeremy Morton (Jez)
First up, Bayes can not work without training - you have to at least
provide a minimum number of samples of what spam and ham look like.
My Bayes scores it at 50%, which simply means it doesn't really look
much like any of the ham or spam that passes through my server. It still
hits 3 DNSBLs for me so would have easily been blocked. Bayes did
exactly what it should have - gave it no score and left other rules to
do their job.
