To enable the spam engine, ensure that Enable antipam engine is ticked. If it isn't, all antispam features are disabled.
The Bayesian filtering works best with similar amounts of ham (legitimate mail) and spam. If the option Automatically balance good and bad email database is enabled, then Rules Manager will ensure that the balance is maintained. If more ham than spam is in the database, any spam messages will automatically be added to the database until the imbalance is removed. If there is more spam than ham, then Rules Manager adds messages that are processed by a rule to the good database until the counts match. Messages processed by a rule are assumed to be legitimate.
By default, the Junk Suspects folder is not enabled. If you tick this option, then Rules Manager will move any mail it is unsure about (i.e. scores below the minimum spam probability, but above the maximum ham probability) will be moved to the Junk Suspects folder (this will be created if it doesn't exist).
If a message is found to contain a link to a known spammer's website, then enabling If message contains spam links, mark as definite spam will mark it immediately as spam. If this option isn't selected, then the percentage of spam links to none spam links is calculated and added to the existing spam score.
Use existing spam analysis in message headers allows you to specify that Rules Manager read the message headers for existing antispam categorisation. If the message doesn't contain these headers, or that they say the message isn't spam, then the message is still filtered using Rules Manager's antispam engine.