'Low' and 'high' spam score thresholds?

[Hirsh, Joshua]

> What in general are other people using?

 At the moment, my low spam threshold is at 5 and is still being delivered
to my users. My high spam is at 20, which is delete from the server. (Yay
for default numbers)

 I've been keeping a close eye (as much as possible, way too much spam to
read it all) and we've been picking up most spam with almost no false
positives. The Bayes DB plays a big role in this as it constantly picks up a
large portion of the spam. Roughly 85% of all spam we received yesterday was
tagged as either BAYES_90 or BAYES_99. DCC and Razor2 also help push the
spam above the threshold, but only tagged 75% and 50% respectively.

 I'm considering lowering my high scoring spam even lower, but we'll see
what happens.

 I had to whitelist a few stupid newsletters (which looked like spam to me,
but apparently my users wanted it) when I first started out, and I haven't
heard of any complaints since then.



 Cheers,
-Joshua