I'm using a simple algorithm based upon a measure of price momentum that I cooked up (volume occasionally enters the picture as well). The method for determining potential trades is mechanical, but, I never built sizing & risk management into the model - that remains discretionary. So, the methodology demonstrates a sound historical basis, but hasn't been 'formally' backtested.
An error was made in the ROI calculation for 5/7/09.
The proper figure is +1.91%, not +1.73%. I had forgotten to enter the purchase of SNPIX on the 6th.