On Wed, 2008-08-06 at 20:25 -0400, Sahil Tandon wrote:
> Karsten Br?ckelmann <[EMAIL PROTECTED]> wrote:
>
> > Hmm... Sahil, Nitin -- guys, you are seriously confusing me.
>
> I am perplexed by your confusion, but I will try to help you.
My confusion stems from different, almost random results all over the
place. That, or you guys have been talking about one spam, but posted
results of another. ;)
(To remind you: One single piece of spam. Three different results of
static RE rules.)
> > Sahil, this is just odd. The examples *do* have the HB_SEP blank line. I
> > guess your download broke or something, but these rules don't apply to
> > the given spamples.
>
> I have no idea re: HB*; perhaps as you suggest, something did "break"
> during the wget.
>
> > Even worse, your rules hit account for a total score of 7.032. Might I
> > ask which rules scores you changed?
>
> What do you even mean by "worse"? People tweak rules in local.cf. To
> satiate your curiosity (sorry for the wrapping):
Sorry, you are right of course. That word sneaked in, because I had a
gut feeling...
> X-Spam-Report:
> * 2.0 BAYES_50 BODY: Bayesian spam probability is 40 to 60%
> * [score: 0.5001]
Bingo! That's *exactly* what I guessed...
That is not a smart move, IMHO. A Bayes score of 0.5 does NOT mean,
Bayes is 50% certain it's spam. It DOES mean, that Bayes does know
nothing. Absolutely nothing.
Between BAYES_00 (aka ~100% sure it is ham) and BAYES_99 (aka ~100% sure
it is spam), BAYES_50 is like a shrugging. It is not a sign of being
spammy. You could just as well lower your spam threshold to 3.0.
If you really feel a need to punish BAYES_50 like *that*, my advice is
to properly train your Bayes instead.
guenther
--
char *t="[EMAIL PROTECTED]";
main(){ char h,m=h=*t++,*x=t+2*h,c,i,l=*x,s=0; for (i=0;i<l;i++){ i%8? c<<=1:
(c=*++x); c&128 && (s+=h); if (!(h>>=1)||!t[s+h]){ putchar(t[s]);h=m;s=0; }}}
