Training yellow messages as spam

[Leisureguy]

When I receive a notification that SpamSieve is unsure that a message is spam, I have been training the message as spam, thinking that this would make SpamSieve more certain on the next such message.

However, these messages are in the Spam folder, and colored yellow. I have just read a response from a developer that training SpamSieve that yellow messages are spam degrades its ability to distinguish good from spam, so now I'm concerned that I've been undermining its accuracy.

a. Should I avoid doing any training of SpamSieve on yellow spam messages?
b. How should I help SpamSieve know that borderline cases (which it has identified as spam) are indeed spam?
c. Should no training be done of messages in the correct folder (spam folder or good folder)?
d. If I've been doing this superfluous (and possibly counterproductive) training, should I zero out training to date and start anew, perhaps letting a bunch of real spam build up in the spam folder for an initial training session?

[Michael Tsai]

a. Should I avoid doing any training of SpamSieve on yellow spam messages?

Yes. SpamSieve will automatically decide how “interesting” it thinks the message is and possibly auto-train itself.

b. How should I help SpamSieve know that borderline cases (which it has identified as spam) are indeed spam?

SpamSieve already thinks those messages are spam. It will assume it’s right unless you train them as good.

c. Should no training be done of messages in the correct folder (spam folder or good folder)?

Right. The exception is when you first install SpamSieve and you’re doing the initial training with some old messages.

d. If I've been doing this superfluous (and possibly counterproductive) training, should I zero out training to date and start anew, perhaps letting a bunch of real spam build up in the spam folder for an initial training session?

No, it’s probably not that big a deal.