I don't think you can apply group filters very easily. Especially not Bayesian filters. As you said in different words, one man's spam is another man's sirloin. Spam filters are fast becoming a requirement, like firewalls and antivirus. Users who install there own Bayes filters can find a ton of relief though. I have been running a Bayes filter for 3 or 4 years now. The accuracy is 99.5% for this past year. In that time I've had 10 false positives that I caught. Most errors are where spams got through, or it was too close to tell. Of those I had 31 falsely classsified messages. Not bad considering I didn't have to personally delete 5,365 spam messages out of my 1,655 legitimate messages.

I'm guessing I probably missed a couple of False Positives, but that's the price required to not have to look at every single piece of spam.

Steve G.