The first number is the sum of all spams (ie the total score of all messages
from that sender), the second score is the number of messages seen.  Divide the
first by the second to get average spam score for that sender.  The AWL process
works by first calculating the regular score for a message, and then doing

finalscore = unmodifiedscore + longtermaveragescore
             --------------------------------------
                             2

I'm not sure if 2 is the right number (haven't played with optimizing it), but
it seems to work pretty good.  Actually, you can come up with a much much better
shrinkage factor, but I'm just too lazy to go pick up my stats books.  Maybe I
will some day.

C

Larry Rosenman wrote:

LR> On Wed, 2002-05-01 at 13:48, Craig R Hughes wrote:
LR> > Apply the attached patch to the 2.20 check_whitelist, or else just grab the one
LR> > I just checked into CVS.  There were more problems which I've fixed than just
LR> > the ones in my initial post.
LR>
LR> Much better.  Thank You.  (now, what do the numbers mean :-).
LR>


_______________________________________________________________

Have big pipes? SourceForge.net is looking for download mirrors. We supply
the hardware. You get the recognition. Email Us: [EMAIL PROTECTED]
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk

Reply via email to