I love crm114. It does an incredible job at filtering all my nasty evil spam — and it rarely screws up. I reclassify, on average, 3 mails a week.

However, I really don’t like the ‘flow interruption’ that I get when a new spam message hits my spam inbox, cluttering it up with that unread count. However, the messages most likely to be missed by my filters are those I most want to get — odd one-offs from AOL or Hotmail — old friends, or business contacts.

I wanted to come up with a sane threshold at which I could ‘mark as read’ inbound spam email. To visualize the scoring CRM114 was coming up with, I made the following charts:

http://serialized.net/crm114/

They catalog, in the top chart, the scores of all the spams that I’ve considered spam (i.e. marked as read and not reclassified) vs all of those that I’ve had to reclassify.

It appears that if I set my cutoff to, say, -70 and below, I’ll have to check my ‘confirmed spam’ every so often, and let the really egrigeous and obvious spam blow into my spambox with nary a shred of notice on my part. We’ll see how that goes. Now I just need a good way to do numerical comparisons in .procmailrc ;)

Dec 14, 02:34 AM − Posted in Tech