TO create a tradeable forecast, takes a lot of time and effort to get it right, I used to do it, but the data kept was expensive, and i didn't quite get the mest data sets. . . you can search on here for some of my forecasts using these methods here is a sample sportsguy
There were two pieces to the forecast, the earnings forecast and then the P/E forecast, then the two were multiplied together. The pink is a 13 week ahead forecast, so it was directionally correct. I also published the same forecasting method in Stocks and Commodities in September 1991, with a pretty good forecast for back then, which I should have perfected, but after grad school and that recession, and the internet consisted of bbs, i got a regular finance job. I recently upgraded to the latest forecasting software version, but i don't have all the data to redo another forecast. .. sportsguy
Linear Regression (method of least squares) is a good tool. How can it not be? It was invented by the greatest mathematician of all time to predict the orbit of Ceres. My preference is to use the time varying generalization of LR, the Kalman Filter. This way, you don't need to do windowing and the problems (in the frequency domain) that this windowing entails.
Here are 3 distributions for the last 1000 daily bars on SPY for: close minus open close minus cubic spline close minus pivot The cubic distribution is not normal, I'm not sure why that is. The best curve for a mean reversion trading approach looks to be the close minus pivot. Maestro, any thoughts on this before I test some trading rules? Tom
Google "breaking vegas" Ron Harris. Now don't get any ideas. Simply an exercise in comprehending the pitfalls of pseudo-random generation via digital algorithms.
Here's a sample equity curve for 1000 daily bars of SPY. Strategy: long on a close below pivot, short above. Also, here is an interesting paper by John Ehlers on deriving a trading strategy from the distribution curve: http://www.mesasoftware.com/technicalpapers.htm (first paper) Tom