http://www.elitetrader.com/vb/showthread.php?s=&postid=914135#post914135 apparently this system was already developed since dec 05
davewtrader : very funny post lol. (i feel ashame laughing at that joke though ) I saw your thread on NuclearPhynance where you said your "system" (or so to speak) takes in 5 mins of past data and forecasts the next 1 tick. Is that right? Two thoughts in my mind right now: 1) If you can obtained a better than 50% accuracy on whether the next tick is up or down you can essentially scalp profitably. (buy the bid, sell the ask) 2) Perhaps there are ways to profit that does not involve forecasting of prices but instead taking advantage of the characteristics of price movement. (this is my personal research) Any links or books to recommend if I want to read up on Mandelbrot's finance related work? (I'm a newbie in Quantitative Finance )
Well, it uses quotes, not ticks and it doesn't explicitly forecast the individual quotes because they are too noisy, but the accuracy is quite high. I measure forcasts of this nature in decibels because it is much more useful than Mean-square-error or R-Squared because both of those measures show near perfect quality (as expected when forecasting smooth wave-like behavior). Decibels are much more useful when calculating signal vs noise (in this case forecast vs actual). You don't really need to know Mandelbrot's stuff in detail for it to be useful, and you probably won't find anything in it directly applicable to making money, but its good for theoretical foundation and creative ways of thinking about problems. I'm not by any means a quant, I just plug stuff together until it works. I've used everything from classical economic theories, chaos theory, signal processing, statistics, computational intelligence, filtering, etc.