On Thu, 28 Feb 2002, Martin Pool wrote:
>
> (I've only been using SA for a little while, so you can take this
> with a bit of salt.)
i just started using it too. just switched to from 2.01 to 2.1 though.
> It really does seem to me like the weightings are less useful in 2.1
> than in 2.01. Some of them are clearly too large, and some of the
> negative weights indicate that the algorithm is "overtrained" and
> looking for patterns when there is not sufficient evidence to
> decide.
i compared the old and the new scores and these are the results: the
10 biggest negative changes:
old new chng
IN_REP_TO -2.0 -13.5 -11.5
MONSTERHUT 1.0 -8.3 -9.3
PGP_SIGNATURE -5.0 -13.4 -8.4
GAPPY_TEXT 2.5 -3.7 -6.2
LARGE_HEX 0.2 -5.9 -6.1
PORN_8 0.5 -5.5 -5.9
BE_AMAZED 1.0 -4.6 -5.6
TRACKER_ID 0.7 -4.9 -5.6
25FREEMEGS_URL 1.0 -4.6 -5.6
DEAR_SOMEBODY 1.0 -4.4 -5.4
the 10 biggest positive changes:
PORN_1 2.9 21.7 18.8
FREE_CONSULTATION 2.3 15.3 12.9
EXCUSE_5 2.2 13.4 11.2
SUBJ_2_CREDIT 1.0 10.9 9.9
PARA_A_2_C_OF_1618 3.0 12.8 9.8
UNSUB_PAGE 1.2 10.8 9.6
SPAM_FORM_RETURN 1.0 10.3 9.3
X_PMFLAGS_PRESENT 1.0 10.2 9.2
YR_MEMBERSHIP_EXCH 0.0 8.8 8.8
MORTGAGE_RATES 2.8 11.3 8.6
i think i might stick with the old rules for now. i run mass-check on
mailboxes first though.
--
Christof Damian
Technical Director, guideguide ltd.
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
