[Rspamd-Users] How to improve Bayes effectiveness?

[Andrei Goldchleger]

>>
>> rspamc stat
>>
>> tell you?
>>
>> For example
>>
>>
>> rspamc -h 192.168.10.2:11334 stat
>> <output snipped>
>> Statfile: BAYES_SPAM type: redis; length: 0; free blocks: 0; total 
>> blocks: 0; free: 0.00%; learned: 3447; users: 1; languages: 0
>> Statfile: BAYES_HAM type: redis; length: 0; free blocks: 0; total 
>> blocks: 0; free: 0.00%; learned: 4216; users: 1; languages: 0
>>
>> How many learned of each?
>
> I get the following:
>
> Statfile: BAYES_SPAM type: redis; length: 0; free blocks: 0; total 
> blocks: 0; free: 0.00%; learned: 16458; users: 1; languages: 0
> Statfile: BAYES_HAM type: redis; length: 0; free blocks: 0; total 
> blocks: 0; free: 0.00%; learned: 46521; users: 1; languages: 0
> Total learns: 62979

Wow, I just found something extremely stupid in my config. I followed 
the configuration suggested in:

https://workaround.org/ispmail/bullseye/filtering-out-spam-with-rspamd/

And a commenter spotted a issue on the sieve scripts that handle 
ham/spam learning - basically it correctly learns spam when the message 
is marked as spam, but learns it as ham when the message is deleted from 
the junk folder (duh). I applied his suggested fix and it seems to solve 
the problem, so hopefully my detection will improve.

Thanks,

Andrei