Adam Denenberg <[EMAIL PROTECTED]> wrote: > [...] > Any recommendations, especially with experience in using the > custom rulesets would be greatly appreciated. Just seeing > which rulesets people have had the most success vs FP ratio > out there.
I'm using the better-known sets recommended in this thread with good success, including regular updates to BIGEVIL TRIPWIRE ANTIDRUG EVILNUMBERS RANDOMVAL. I've also got the last copies of Jennifer's BACKHAIR and WEEDS2 sets in play. I've added some "best of" rules from other sets that DO have a fair incidence of FPs, but I have them scoring very low (if at all). An FP isn't so bad so long as no single rule causes mail to be scored as spam by itself. I then use metas to score on the basis of COMBINATIONS of these rules, with scores that will flag them as spam. I'm generally finding these to be as effective as trying to roll up new rules to definitively match newly detected patterns. These plus bayes have been very solid for me. - Bob
