Although the supercomputer theory is interesting, I find it unrealistic from a programming standpoint. I mean, somebody has to create the complex algorithms that take into account an almost infinite number of variables, which all must be weighed and analyzed in real-time. This will require tons of computing power as well as an army of programmers that will need their hands held by market gurus that can convey exactly what logic needs to be coded. Scary. Earlier, ArchAngel mentioned the arbitratge system that takes advantage of the premium between the S&P futures and the S&P cash index. This is an example of a successful computer-automated trading system, but it's also one that is mathematically simple to implement, as there are only a few primary variables involved. It's a perfect play for a computer-controlled system because the logic is simple, it's fairly easy to program, and it requires relatively low amounts of computing power.