A while ago I replaced my Amavis-Spamassassin-Rules_du_jour-MaiaMailguard setup with a single product called DSPAM.
This anti-spam solution relies completely on statistical spam analysis and takes less than a second to process (spam-detect/virus-detect) a mail on my server. Spamassasin would take anywhere between 10 and 30 seconds to detect spam, and even with all kinds of custom rulesets in place and using bayesian spam detection the accuracy was only around 95%.
After my switch to DSPAM I immediately fed it a spam/ham corpus from my personal archive. This was enough to give it an extremely good accuracy.
These are my current statistics for the filter’s accuracy:
The ‘misses’ you see in the ‘Total processed by filter’ statistics are caused by the way I fed the spam/ham corpus into my clean DSPAM installation. After the training I’ve reset the filter, and as you can see my spam-catch rate is now 100%
DSPAM really is an excellent way to rid of spam!