Hello,

lately, I am having problems with spam mails being marked as good (e.g. Bayesian anaylsis gives a 96% that the mail is good), but when I run them throught the spam simulator it says "Bayesian analysis suggests there is a 96% chance this message is junk"

Any idea, what is going on ? Is there a problem with Bayesian that the good and bad percentage is mixed up ?

Thanks,

Oliver

Oliver,

Is this happening with EVERY message or just a handful? If this is every message, please contact our support department as some further investigation will be needed. 

If this is happening with just a handful of messages, are you sure no one marked the message as junk? How often does this happen?

An FYI: Bayesian DB is stored in flat files in the config folder. The file names look like HamWords_XXX.dat and SpamWords_XXX.dat. If you think Bayesian has somehow gotten corrupted, you can remove these files and configure your Basyesian run in "learning" mode. This will reset the Bayesian on your end.