The Spam Score is a statistical probability calculated by the Bayesian filter (this is different from how it worked in Spamfire 1.6). The Spam Score for each message is shown in the left-most column of the Spamfire Mail Window.
The spam score for a message ranges between 1 and 99. A score closer to 1 indicates a good message. A score closer to 99 indicates a spam message. A score of 50 indicates an unknown message.
When you first start using Spamfire, you will see more messages with scores of 50. As time goes on, Spamfire will get better about recognizing email and you will see more scores at the extreme ends of the scale. Spamfire can train itself, and you can also improve its accuracy by correcting Spamfire when it makes a mistake.
More about what the different scores mean.