[Spambayes] Msg class broken?

[T. Alexander Popiel]

In message:  <006c01c2d011$acf230d0$530f8490 at eden>
             "Mark Hammond" <mhammond at skippinet.com.au> writes:
>[Alex]
>> On the other hand, I'm now having a small bit of trouble with
>> my graphs;
>> spambayes is too darn good.  On my real data, in less than two days
>> training, the error rate is in the noise floor.  I haven't tried all
>> the different training methods discussed yet, but it's kinda hard to
>> differentiate when the filter just doesn't screw up.
>
>Can you post this anyway?  I'm particularly interested in this, as I feel it
>is the roughest edge left on the outlook client - that initial training
>period.

Certainly, once I'm done with tightening up the presentation a bit.
I'm also going to try it with significantly smaller and more balanced
amounts of data; I suspect most people don't have a 2:1 spam:ham ratio.

- Alex