Re: a way to set scores for new rules in updates

[Justin Mason <jm...@jmason.org>]

If it did take overlap into account, I'd be curious.  I thought it
didn't...

--j.

Loren Wilton writes:
> Bob had a technique that worked reasonably well, and only required some 
> minor human thought to usually come up with good numbers.  I'm pretty sure 
> that it took overlap into account.  (Although the overlap runs may have been 
> something he did manually, I don't recall.)
> 
> He seems to be gone to other places now, but I'm pretty sure he left a 
> description of how his masscheck and score generation worked someplace 
> around here, probably in the archives of the newsgroup.
> 
> I think it would be worth looking into, since it didn't require much more 
> than some slightly modified masscheck tools, as best I recall.
> 
>         Loren

Re: a way to set scores for new rules in updates

[Doc Schneider <ma...@maddoc.net>]

Justin Mason wrote:
> If it did take overlap into account, I'd be curious.  I thought it
> didn't...
> 
> --j.
> 
> Loren Wilton writes:
>> Bob had a technique that worked reasonably well, and only required some 
>> minor human thought to usually come up with good numbers.  I'm pretty sure 
>> that it took overlap into account.  (Although the overlap runs may have been 
>> something he did manually, I don't recall.)
>>
>> He seems to be gone to other places now, but I'm pretty sure he left a 
>> description of how his masscheck and score generation worked someplace 
>> around here, probably in the archives of the newsgroup.
>>
>> I think it would be worth looking into, since it didn't require much more 
>> than some slightly modified masscheck tools, as best I recall.
>>
>>         Loren

I have all the notes on how he (Bob) did his scoring. I can forward 
these on to you Justin if you want.
I even wrote a small script where you add the ham and spam totals into 
it and it then spits out a fairly good score. This was written using 
Bob's scoring algorythm(sic -- I need more coffee).

His mass-checker script does have an option for --overlap which we in 
SARE do use to check for overlaps for our own rules and also the SA rules.

Let me know and I'll share all of this with you. 8*)

-- 

  -Doc

  SA/SARE -- Ninja
    7:36am  up 3 days, 23:02, 17 users,  load average: 1.12, 0.70, 0.52

  SARE HQ  http://www.rulesemporium.com/