[Mimedefang] typical spamassassin results

[David F. Skoll]

On Thu, 24 Jun 2004, Rudolph Pereira wrote:

> I am trying to work out "typical" results for how much spam SA catches
> in it's default configuration, without bayes or dnsrbl (or any network
> tests) configured. My understanding was that it would be in the order of
> 70%, but I am getting between 30-40% on the corpus I have (~2k spam all
> from the last few months).

What version of SA do you run?  I find it's highly variable; some people
get 80% coverage with stock SA, and others only get 25%.

I strongly recommend the SURBL rules, and some of the other rules from
http://www.rulesemporium.com/, and the "popcorn", "weeds" and
"backhair" rules at http://www.spamfighter.org/bb2/download.php?list.6

With all of the rules, plus greylisting, plus some HELO checks,
plus Bayesian filtering, I get maybe one spam a month that makes
it through.

Regards,

David.