I watched an excellent programme last night called Derren Brown Investigates: The Ghosthunter. It was interesting for two reasons. First, the best evidence for the existence of ghosts that a dedicated believer had amassed in many years of research was extremely lame. Second, the man was totally convinced by his evidence.
This made me think of technical analysis. Lots of things (like Fibonacci Retracements) might often seem to work, but you have to ask yourself, 1. is there any reason to think that they should and 2. is the belief that they work based on any real (i.e. quantitative) evidence?