Some rules of thumb:
- add some credit to all e-mail at the start. It lets you decrease scoring later - based on further rules - down to zero, what is a sign of spam; Some people add 2, some 100 at the beginning
- with the rule above in mind, add set score_threshold_delete=0
- choose one of the two scenarios: you add points after parsing all rules and based on achieved scoring you want Mutt to "F"lag a message, or you "F"lag a message manually what adds some points to a message (what brings it higher when sorting by score). In the first case add set score_threshold_flag=15, in the latter score "~F" +5
- if you have a lot of e-mails just as a reference/documentation, and do not read it manually, you can find it useful: set score_threshold_read=5
- use George Boole algebra: score "~s Apple | ~s apple" +5
- increase scoring for e-mails you want (score "~f firstname.lastname@example.org" +666) and punish mails you hate (score "~s training | ~s Certificate" -100). After that (and score_threshold_delete set) all you need to remove spam is to press '$'.
- make obsolete obsolete: score "~d >1m" =2
color index brightcyan default "~n 5-15"
color index brightmagenta default "~n 16-100"
You can always score manually executing the commands above in Mutt after a colon:
:score "~f email@example.com" +10
and then sort by pressing 'o'.