Bayes again

[Pete]

I have batyes set up and working, but we only use autolearning - no
manual learning. I was reading through the archives and some have said
not to use bayes at all if i am not using manual leanring? I have found
over the past month that more and more spam is getting through, all with
negative scores

I saw Julians recent post on his setup and have decided to try similar,
I will install DCC and razor2 and add the RBLs (to SA), i have just
added bigevil and will then upgrade SA to 2.61. I am hoping this mean
our system is more fire and forget - we use high spam scores to stop
spam, we use no tagging features.

Any tips on bayes? Should i stop using it in this configuration and use
the other tools to try and get this system to chug along with less
administritive effort?

Currently have postfix 2.016 red hat 9 MS 4.25-4 and SA 2.6