While modifying the GA, I noticed that I'd run the scoring for 2.1 with a population of only 20 -- normally I run with about 100 but had lowered it or speed while doing some testing. The small population size almost certainly had a major negative effect on the scores produced. I'm rerunning now with the limits on individual scores mentioned before, plus the bigger population size, and so far getting much, much better results in terms of number of false identifications while staying within more same scoring limits for the individual rules.
C _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
