should I change score for Bayes_90?

[Mark Nienberg]

I guess you are using an older version of SpamAssassin, because the scores from
the final release of v2.6 are as follows:

score BAYES_70 0 0 2.142 2.255
score BAYES_80 0 0 2.442 1.657
score BAYES_90 0 0 2.454 2.101
score BAYES_99 0 0 5.400 5.400

Also, I don't see why you would have higher scores for lower bayes probablilities, the
way your third column does.  (Whoops, I just noticed that the official scores do the
same thing for BAYES_80 in the fourth column!) Of course, if you are using network
checks in addition to bayes, then only the fourth column matters.

I've only been using the final 2.6 for a few days, but it seems like the BAYES_99 is
showing up a lot more often than it did in the early beta release of 2.6 that I was
using.
---
Mark

On 21 Oct 2003 at 9:02, Richard Ahlquist wrote:

> I too have adjusted my Bayes scores. Here they are.
>
> score BAYES_70 0 0 5.593 3.310
> score BAYES_80 0 0 5.300 3.862
> score BAYES_90 0 0 4.900 4.200
> score BAYES_99 0 0 5.200 4.900
>
> auto_learn 1
> auto_learn_threshold_spam 10
> auto_learn_threshold_nonspam -3.0
>
> Hopefully I got those correct. I receive nearly 90% spam on the accounts
> affected by this mails server (some of the addresses were used for newsgroup
> postings as far back as 94). I know I have everything set really low but it
> works for me. Spam threshold is 5, Bounce is set to 7. I have gotten only
> one false positive in the last few months (I monitor the subject and authors
> of bounced messages using MailWatch).
>
> -----Original Message-----
> From: MailScanner mailing list [mailto:MAILSCANNER at JISCMAIL.AC.UK] On Behalf
> Of Chris Yuzik
> Sent: Monday, October 20, 2003 11:58 AM
> To: MAILSCANNER at JISCMAIL.AC.UK
> Subject: Re: should I change score for Bayes_90?
>
> On October 20, 2003 03:40 pm, Chris.Campbell at fac.com wrote:
> > I have a similar issue... and I was wondering if anyone responded on this
> > thread... I would like to know if others are changing the bayes* scores...
> > I know I have wrote about 100 custom rules, and changed about 50
> > points.....but i haven't touched the bayes yet... mostly because I haven't
> > found
> > a good way to use sa-learn on my gateway mailrouter boxes yet... (internal
> > users are  lotus notes users, and it doesn't support forward as attachment
>
> Chris,
>
> Nobody responded to the thread (exept you). I analyzed a bunch of spam and
> decided to change the scores myself. I arbitrarily decided to set Bayes_90
> to
> 4.25 and Bayes_99 to 4.50. So far, it seems to be working well.
>
> Chris