Rules Manager incorporates antispam protection, and uses several techniques to identify spam. By default antispam is turned off - you will need to enable it in the options.
One of the techniques used to identify spam is called Bayesian analysis. This works by marking messages as either spam or ham (legitimate mail), and RM then builds up a database of these messages. This database is used to analyse messages to determine a spam score for a message.
Bayesian analysis needs a certain amount of data to be effective. By default, the database will have no data, as no messages have been classified. You can either build a database using pre-sorted mail (i.e. mail that has been sorted into folders for spam and legitimate mail), or elect to build up the database as you use Outlook. Until Rules Manager has enough data for analysis, the Bayesian filter will be disabled. If the "Use Junk Suspects" folder is enabled, this means that ALL messages that aren't identified as spam by other means will be moved to Junk Suspects. You then need to identify these messages as spam or legitimate mail.
The minimum amount of data required for Bayesian analysis is ten of each item (i.e. ten good messages, ten spam messages). Any less than this and the comparisons are essentially meaningless.
The advantage of Bayesian analysis is that it is customised to your mail system. However, other antispam technique are incorporated that work on a more general level, and in practise these should catch at least 75% of spam messages.