There are two basic methods of SPAM detection algorithms Heuristic and Bayesian.
Heuristic is a fixed algorithm that guesses at the email if it is SPAM. It is not updatable. Once a Heuristic algorithm is defeated it becomes worthless in detecting SPAM.
Bayesian has to learn over time what it SPAM and what is not SPAM, but is adaptable and updatable. Spammers are constantly changing the way they send spam such as v1agr@. They have 1.000’s of new ways to try to get around detection.
Hotmail is a good example. They have deployed a new Heuristic SPAM filter that that completely stops all the junk from getting into my inbox. Well that lasted for about 2 months. Now I still don’t get the old junk mail, but because a spammer has figured out how to defeat it, I am getting 30 identical SPAM emails a day on that account.
Bayesian is better as it allows you to adapt as the spammers adapt.