SpamSieve’s filtering accuracy depends on proper training, the most important part of which is training all the mistakes. If you lead SpamSieve to think that a spam message is good, it will think you want to see more messages like that. The more prompt and consistent you are at training, the less of it you’ll actually have to do.
If there’s been a lapse in training the mistakes, e.g. because of a setup problem or because you were temporarily away from your Mac or overwhelmed by spam, SpamSieve has some tools so that you can fix the training without having to reset and start over. Here are some different areas to check:
Find any spam messages in the inbox in your mail program and train them as spam.
If there were any spam messages in your inbox that you deleted from your iPhone instead of training, find them in the trash and train them as spam. (In the future, if you see spams in the inbox when you’re away from your Mac, move them to the special TrainSpam mailbox instead of deleting them.)
Check the Good Messages and Spam Messages tabs of the Corpus window. If you see any messages that are in the wrong section, you can fix them by training from that window.
Check the Allowlist window, sorted to show recently created rules. If you see any spammy names or addresses, they should already be disabled. If there are spammy rules that are enabled, don’t disable them. These enabled rules are a sign that there are spam messages that were not trained as spam. When you fix the training of those messages (as described below), SpamSieve will automatically disable the rules.
When using the Auto-train as needed option, some messages with cause the blocklist and allowlist to be trained without the messages being added to the corpus. Thus, they cannot be fixed as described in Step 3. To fix these, you can go through the Log window, starting at the bottom (the newest messages). If you see any spam messages that say Predicted: Good in green, that means SpamSieve thinks it was correct about the message being good. You should train these messages as spam. Likewise, if you see any good messages that say Predicted: Spam in green, train them as good.
It may not be not practical to go through all the old messages in the log as described in Step 5. If there’s a particular spam message that SpamSieve thought was good, you can look into why and then focus on correcting the messages that led to SpamSieve not catching that spam. In the Log window, find the Predicted: Good entry for the message:
There’s more information about this in the examples in the Searching section of the manual.
In following Steps 5–6, you may find that there are messages in the log that SpamSieve won’t let you train. It is only possible to train messages in this way if SpamSieve is storing the full message data. Normally, it does this for messages that were received in the last 30 days, and this can be adjusted using the Prune full message data in log setting. Once you’ve fixed the training of the available messages, you can go back to the allowlist (Step 4) and disable any spammy rules that remain.
Resetting SpamSieve
If it seems like there are too many old messages for you to go through and fix, the other option is to start over with a clean slate. To reset all of of SpamSieve’s training:
Quit SpamSieve and your mail program.
Open the Library folder and drag the SpamSieve folder to the trash:
/Users/<username>/Library/Application Support/SpamSieve/
Follow the instructions in the Do an Initial Training section.