Change the score of BAYES_9*

[The Doctor <do...@doctor.nl2k.ab.ca>]

I wish to make a system-wide change for BAYES_95 and BAYES_99 to 
score 1000.0 .  999.999% of those e-mail scoringthat high 
are worthy of GTUBE status.

How can make that change systemwide?

-- 
Member - Liberal International	
This is doctor@nl2k.ab.ca	Ici doctor@nl2k.ab.ca
God, Queen and country! Beware Anti-Christ rising! 
Birthdate: 29 Jan 1969 Redhill Surrey England

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.

Re: Change the score of BAYES_9*

[Ned Slider <ne...@unixmail.co.uk>]

Matt Kettler wrote:
> The Doctor wrote:
>> I wish to make a system-wide change for BAYES_95 and BAYES_99 to 
>> score 1000.0 .  999.999% of those e-mail scoringthat high 
>> are worthy of GTUBE status.
>>
>> How can make that change systemwide?
>>   
> in local.cf add:
> 
> score BAYES_95 1000.0
> score BAYES_99 1000.0
> 
> If you use spamd or an API level tool that caches a Mail::SpamAssassin
> object (ie: MailScanner), it will need to be restarted to read the new
> config.
> 
> However, I will warn you this is a bit dangerous. Theoretically, the
> false positive rate of those two should be 5%. (ie: 5% of the mail they
> match is nonspam mail).
> 
> That said, I also don't understand why such a strong score. That's
> higher than a manual whitelisting will compensate for (-100). Do you
> really want this to be so high that it over-rides your explicit
> whitelists? Why not use something like 20 or 50?
> 
> GTUBE is scored so high because it needs to over-ride any whitelisting.
> But nothing else should ever need such a high score.
> 
> 

I agree that in the case of a well trained Bayes system it's useful to 
increase the scores for BAYES_99 and BAYES_95 (and likewise maybe 
increase the negative scores for BAYES_00 and BAYES_05), but I typically 
set these to around the spam threshold score so hits against them should 
have the mail marked as spam but there is also scope for hits against 
any negatively scoring rules to correct against any FP hits on Bayes 
(which in my experience are extremely rare - YMMV).

As Matt says, I see little advantage to setting Bayes scoring much above 
the spam threshold.

Re: Change the score of BAYES_9*

[Matt Kettler <mk...@verizon.net>]

The Doctor wrote:
> I wish to make a system-wide change for BAYES_95 and BAYES_99 to 
> score 1000.0 .  999.999% of those e-mail scoringthat high 
> are worthy of GTUBE status.
>
> How can make that change systemwide?
>   
in local.cf add:

score BAYES_95 1000.0
score BAYES_99 1000.0

If you use spamd or an API level tool that caches a Mail::SpamAssassin
object (ie: MailScanner), it will need to be restarted to read the new
config.

However, I will warn you this is a bit dangerous. Theoretically, the
false positive rate of those two should be 5%. (ie: 5% of the mail they
match is nonspam mail).

That said, I also don't understand why such a strong score. That's
higher than a manual whitelisting will compensate for (-100). Do you
really want this to be so high that it over-rides your explicit
whitelists? Why not use something like 20 or 50?

GTUBE is scored so high because it needs to over-ride any whitelisting.
But nothing else should ever need such a high score.