On Wed, Sep 10, 2003 at 10:36:33AM -0400, Matt Kettler wrote: > At 03:55 PM 9/10/03 +0200, Carlo Wood wrote: > >The attached mail got through spam-assassin without a problem... > >Why? Where is the BAYES_* test? > >Does this mean that the bayes engine thinks this is less than 10% > >chance to be spam? That would be ridiculous! > >How can I test on which keywords it is basing that this ham? > > 1) do you normaly get bayes matches? (ie: have you trained at least 200 > spam and 200 nonspam messages?)
Yes I did. I've taught it thousands and thousands of both (of the order of 6000 of both I think). However... I did the debug trick of point 3) and it says: debug: debug: Only 81 ham(s) in Bayes DB < 200 *cry* My database got screwed up! This is very very serious wrong :/ -- Carlo Wood <[EMAIL PROTECTED]> ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk