I don't understand all the dynamics, but I would have thought that a BAYES_80 with the FORGED_HOTMAIL_RCVD would be sufficient for a higher score. My mistake.
That's ok.. the real dynamics are actually very complex. Even the spam vs nonspam hitrate comment I made is really a gross simplification of what the dynamics are like.
The actual dynamic of score assignment is based on real-world combinations of hits. So, to figure out the proper score for a given rule, you need to not only consider that rule and how often it hits, but also every combination of other rules that also fire off at the same time, and the hitrates of those rules, etc.
This kind of massive interaction between the rules is why a GA is used in the first place.. It's just way too hard to crunch through it all by hand.
However, there is a strong correlation between the score of a rule and it's spam vs nonspam hitrate (sometimes called S/O for short) and that's a much simpler thing to look at as a first-check. However, for some rules even that simple correlation doesn't hold true.
------------------------------------------------------- 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
