SpamSieve is nearly 100% accurate, but only when properly trained. Here’s how you can check the training:
The most common problem is not training all the mistakes, so that SpamSieve thinks some good messagse are spam and/or some spam messages are good. The Correct All Mistakes section describes how to properly train the mistakes. The Fixing Uncorrected Mistakes section describes how you can confirm that you have been doing this and fix the training if necessary.
Choose Statistics from the Window menu. Does the Corpus section show at least 100 spam messages and 100 good messages? If not, follow the instructions in the Do an Initial Training section.
Check the percentage of spam messages in the Corpus section of the Statistics window. It should be between 55% and 75%. If not, train SpamSieve with more spam or good messages so that the percentage is in this range.
Check the number of spam and good messages in the Corpus section of the Statistics window. If the corpus is very large (more than 4,000 messages) and SpamSieve is not catching your spam, the corpus may contain incorrect or obsolete messages that reduce SpamSieve’s effectiveness and make it slower to adapt. Choose Reset Corpus… from the File menu and then follow the instructions in the Do an Initial Training section.
Note: You should only reset the corpus after you’ve checked all the other items in the Why is SpamSieve not catching my spam? section.