You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by Paul Smicker <sm...@renovis.com> on 2005/01/17 19:13:20 UTC
BAYES_99 = 1.9?
Hi All--
I've upgraded SA from 2.63 to 3.0.2 and I've noticed a big difference
in my bayes scores. Under 2.63 a 99% bayes probability would give me a
score of 5.4, now I get only 1.9:
Content analysis details: (1.9 points, 4.0 required)
____
pts rule name description
---- ----------------------
--------------------------------------------------
1.9 BAYES_99 BODY: Bayesian spam probability is 99 to
100%
[score: 1.0000]
____
This is causing a noticeable amount of spam to get through now. What's
the thinking behind such a low score for mail that has just been given
a 99% chance of being spam ? Will I be asking for trouble if I modify
50_scores.cf to match 2.63 bayes scores?
--Paul
---------------------------------------------------------------------------------------------------------------------
This email may contain material that is confidential and privileged and is for the sole use of the intended recipient. Any review, reliance or distribution by others or forwarding without express permission is strictly prohibited. If you are not the intended recipient, please contact the sender and delete all copies.
Re: BAYES_99 = 1.9?
Posted by Matt Kettler <mk...@evi-inc.com>.
At 08:37 PM 1/18/2005, up@3.am wrote:
>BTW, it looks like bayes_90 has been deprecated. When I run a lint on my
>local.cf, I get:
>
>warning: score set for non-existent rule BAYES_90
Yes, several of the old ranges in 2.64 no longer exist.. For 3.x they
changed the ranging a bit, creating some new ones, merging together others..
For example, BAYES_90 went away, as did 01,10, 30, 44, 56, and 70. However,
95 and 05 were added.
Basically before there was a lot of "splitting of hairs", particularly
near 50, that was largely useless. Bayes 40,44,50 and 56 all had
more-or-less the same score. Zero, or damn close to it. So they made them
all into the new BAYES_50...
It would also appear that they found that 10% away from the extremes (10
and 90) were not as useful as 5% (05 and 95). I suspect this was all done
after some graphing of frequency... In on my experience, the vast majority
of mail winds up in 01, 99 or 50, so it makes sense to coalesce some of the
other ranges...
Re: BAYES_99 = 1.9?
Posted by up...@3.am.
On Mon, 17 Jan 2005, Thomas Arend wrote:
> With network test enabled bayes scores lower. This is a problem when the
> network test don't fire when the spammer uses a new server. Therefore I have
> raised the bayes scores for bayes_99. I seldom get bayes_90 so I didn't raise
> the scores for bayes_90.
Rational, I suppose, but I use the network tests and still found it
neccessary to bump the bayes 9x up to get decent results after upgrading
from 2.63 the other day.
BTW, it looks like bayes_90 has been deprecated. When I run a lint on my
local.cf, I get:
warning: score set for non-existent rule BAYES_90
James Smallacombe PlantageNet, Inc. CEO and Janitor
up@3.am http://3.am
=========================================================================
Re: BAYES_99 = 1.9?
Posted by Thomas Arend <ml...@arend-whv.info>.
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Am Montag, 17. Januar 2005 19:56 schrieb John Fleming:
> > The Bayes_9x scores were a bit low for our needs. So we bumped them
> > up. That is one of the great things I like about SA. You can customize
> > it to meet your needs.
> >
> > Shane
>
> My part of the thread was that the scores don't consistently go -up- with
> increased percentage - see previous. tnx
This has been discussed before.
In short the scores are calculated with a mass check. Thats the reason why
bayes_90 is cored higher.
With network test enabled bayes scores lower. This is a problem when the
network test don't fire when the spammer uses a new server. Therefore I have
raised the bayes scores for bayes_99. I seldom get bayes_90 so I didn't raise
the scores for bayes_90.
Thomas
- --
icq:133073900
http://www.t-arend.de
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (GNU/Linux)
iD8DBQFB7B0pHe2ZLU3NgHsRArzXAJkBe60sgW2+B9Pd9rQ72nrDu7ywFgCeP0RB
/W7fH9ZrpQN3Odbcuk83hHU=
=7dzL
-----END PGP SIGNATURE-----
Re: BAYES_99 = 1.9?
Posted by John Fleming <jo...@wa9als.com>.
> The Bayes_9x scores were a bit low for our needs. So we bumped them
> up. That is one of the great things I like about SA. You can customize
> it to meet your needs.
>
> Shane
My part of the thread was that the scores don't consistently go -up- with
increased percentage - see previous. tnx
Re: BAYES_99 = 1.9?
Posted by shane mullins <ts...@wise.k12.va.us>.
Hey John,
The Bayes_9x scores were a bit low for our needs. So we bumped them
up. That is one of the great things I like about SA. You can customize
it to meet your needs.
Shane
----- Original Message -----
From: "John Fleming" <jo...@wa9als.com>Subject: Re: BAYES_99 = 1.9?
To: <us...@spamassassin.apache.org>; "Paul Smicker"
<sm...@renovis.com>
Sent: Monday, January 17, 2005 1:31 PM
> > 1.9 BAYES_99 BODY: Bayesian spam probability is 99 to
> 100%
> [score: 1.0000]
> ____
>
> And what's the rationale for 95% scoring lower than 80%?? Sorry if
this has
> been previously mentioned.
>
> score BAYES_00 0 0 -1.665 -2.599
> score BAYES_05 0 0 -0.925 -0.413
> score BAYES_20 0 0 -0.730 -1.951
> score BAYES_40 0 0 -0.276 -1.096
> score BAYES_50 0 0 1.567 0.001
> score BAYES_60 0 0 3.515 0.372
> score BAYES_80 0 0 3.608 2.087
> score BAYES_95 0 0 3.514 2.063
> score BAYES_99 0 0 4.070 1.886
>
Re: BAYES_99 = 1.9?
Posted by John Fleming <jo...@wa9als.com>.
> 1.9 BAYES_99 BODY: Bayesian spam probability is 99 to
100%
[score: 1.0000]
____
And what's the rationale for 95% scoring lower than 80%?? Sorry if this has
been previously mentioned.
score BAYES_00 0 0 -1.665 -2.599
score BAYES_05 0 0 -0.925 -0.413
score BAYES_20 0 0 -0.730 -1.951
score BAYES_40 0 0 -0.276 -1.096
score BAYES_50 0 0 1.567 0.001
score BAYES_60 0 0 3.515 0.372
score BAYES_80 0 0 3.608 2.087
score BAYES_95 0 0 3.514 2.063
score BAYES_99 0 0 4.070 1.886