On Tue, 14 May 2019 10:18:35 -0400 Kris Deugau wrote: > Matus UHLAR - fantomas wrote: > > On 14.05.19 06:18, cyflhn wrote: > >> but what about this one "FREEMAIL_FORGED_REPLYTO". why it got 2.1
> As for the score, it's autogenerated based on statistical analysis of > a large corpus of mail along with most of the other stock rules: > https://wiki.apache.org/spamassassin/NightlyMassCheck My understanding is that scores in 50_scores.cf haven't been optimized for 10 years since the very fast perceptron fell-off and they had to revert to the extremely slow genetic algorithm. The autogenerated scores for a minority of the rules are in 72_scores. This isn't entirely accurate because some rules are scored in both files, but it gives a rough indication of the scale of problem. $ grep -E '^\s*score' 50_scores.cf| wc -l 662 $ grep -E '^\s*score' 72_scores.cf| wc -l 210
