When will MW Pro be fixed???

[lin_kot]

I continue to get unclassified messages ... those that I have classified many-many times ... ie: Consolidate Dept, Bankruptcy Help, Masters Degree, Online Doctorate, Replacement Windows ... SPAM!

Presently FirstAlert is checked but seemingly not working.

[Sidewinder]

Enable the side panel and take a look at what each of the tools has to say about the messages in question, you have to select the message to see the results.
You can also enable the Total and Bayesian columns in the grid to observer the evaluation of each message in the grid without having to select them.
The Bayesian debut tab provides all of the information about the Bayesian evaluation including the words that contribute to the evaluation.
Are you marking the messages as Spam before you do the Wash or just marking them for deletion?

[lin_kot]

Marking as spam before doing the wash. When I look at Blacklist there is nothing listed!

[Sidewinder]

Marking as spam before doing the wash. When I look at Blacklist there is nothing listed!

Blacklist entries are only stored if you right click on the message header and make the selection for blacklisting. You can also adjust the effects of Blacklisting matches by changing the slider in the Spam Tools options. That will help overcome a Bayesian evaluation that makes the message look OK based on content.

There is also the option of creating a Filter based on Spam Tools with a user is on my blacklist. The filter has it's on user selected weighting and options for marking for deletion.

[lin_kot]

I'm a long-time user of MW ... is this something new with this recent version? I have never had to tweak so many settings ... to me the previous version sorted very well on its own!

[Sidewinder]

I'm a long-time user of MW ... is this something new with this recent version? I have never had to tweak so many settings ... to me the previous version sorted very well on its own!

MWP 2010 is a brand new product from the the ground up with significant improvements designed to counter the sophisticated technical efforts of the spammers.
To get a short tutorial regarding the new evaluation process check out this topic:
http://forum.firetrust.com/viewtopic.ph ... 14&start=0

[mkcomputing]

MWP 2010 is a brand new product from the the ground up with significant improvements designed to counter the sophisticated technical efforts of the spammers.

Interesting that you should day that.

Seems like the same old thing with a new interface, and one that I am happy to accept, even though there are still serious defects in the screen writing, for example, along with a few missing features. (Like hiding messages deemed to be OK.)

Rather than sending us off into what might be oblivion, any chance of two or three bullet points on these new, and so far undetected, significant improvements?

The old version rarely delivered rubbish to my inbox. What should I look for to tell me that the new one is doing a better job?

We won't even go into the annual-fee model, which some people suspect is the real reason for the new version.

[lin_kot]

I'm not on board yet either. Seems the new product should have been tested prior to asking us to pay for it and end up being inconvenienced (to say the least) ... and sources of bug reports. I'm pretty discouraged ...

[rusticdog]

I'm not on board yet either. Seems the new product should have been tested prior to asking us to pay for it and end up being inconvenienced (to say the least) ... and sources of bug reports. I'm pretty discouraged ...

As best as I can tell this isn't a case of not being tested, it's a case that MailWasher hasn't yet been trained enough. When you classify your email and Wash Mail, MailWasher learns....as more emails process you will find that MailWasher will get better and better over time.


The old MailWasher has the same issues, in fact I find the old MailWasher when testing from scratch caused quite a few headaches getting emails wrong in those early days. The new version has a more scaled result, so in cases where MailWasher isn't entirely sure it will often mark the emails back as undetermined and effectively asks you to classify the message until it can be sure.


As per the post here it explains how to see this in more detail, I'd be interested to know what your Evaluation Results look like, in particular the Learning which is the topmost one.
http://forum.firetrust.com/viewtopic.ph ... 14&start=0

You can also go View >> Bayesian Debug >> this adds another tab to the preview area which has the full breakdown. It's not in a nice readable format, but may help to look at in conjunction with the above.

Here's an example..

Code: Select all

DECODED_BASE64_SUBJECT: 0
SUBJECT_WORDS: 2
FROM_WORDS: 6
CONTENT_TYPE: multipart/alternative;  boundary="----------D3AE925F8BDA758"
NUM_PARTS: 2
ENCODING: 7bit
TYPE: text/plain; charset=iso-8859-2
BYTES: 773
NUM_RAW_WORDS: 31
ENCODING: 7bit
TYPE: text/html; charset=iso-8859-2
BYTES: 1708
NUM_RAW_WORDS: 34
MULT: 1.155827E-011
COMB: 2.570212E-004
RAWSPAMYNESSE: -9.999999E-001
RAWSPAMYNESS: -1.00000
SPAMYNESSE: -7.499999E-001
SPAMYNESS: -0.75000
GOODCOUNT: 12
BADCOUNT: 4
GOODWORDCOUNT: 6799
GGAIN: 1.000000
BADWORDCOUNT: 1085
BGAIN: 1.000000
INTRESTINGWORDCOUNT: 20
WORDCOUNT: 37
TOTAL_WORDCOUNT_FACTOR: 1.0000
INTERESTING_WORDCOUNT_FACTOR: 0.5000
WORD: asubject_lovemaking 	prob=0.011000       	occurrences=1
WORD: hope                	prob=0.011000       	occurrences=1
WORD: information         	prob=0.011000       	occurrences=3
WORD: quick               	prob=0.011000       	occurrences=1
WORD: range               	prob=0.011000       	occurrences=1
WORD: delivery            	prob=0.949493       	occurrences=2
WORD: products            	prob=0.949493       	occurrences=1
WORD: looking             	prob=0.758056       	occurrences=1
WORD: purchase            	prob=0.258232       	occurrences=3
WORD: service             	prob=0.406663       	occurrences=2