New version

[Boris 'pi' Piwinger]

Greg Louis wrote:

>> I wouldn't call it a straw man, as that implies it is false.  It is not
>> a false case, just a worst case.
> 
> I disagree here -- let's settle on "artificial" rather than "false" --

There is no need to argue either way. It really depends on
parameters. For small min_dev it is pretty unlikely to have
only insignificant tokens. For very big min_dev it might
well happen once in a while.

Actually, due to problems like this I reduce my value (note
that we really have no clue how to optimize parameters for
training to exhaustion, but this tuning seems to be of very
much smaller importance compared to full training).

> I get 300 to 500 spams a day and they're quarantined, not discarded.  I
> scan the quarantine daily but that's not certain to detect every fp. 

That seems to be the main problem with just every filter.
And since I just receive hundreds of bounces (someone sent
spam with my address), those are not recognizable from
Sender and Subject. So far, bogofilter seems to catch all of
them perfectly.

pi