This example takes a different approach to the rule based Trading Strategy used in the QQQQ and Spy Crossover and Detrend examples.  Rather than trying to specify exact trading rules, we created a neural network prediction with the MESA8 Detrend, BandPass, and Crossover Momentum indicators.  This might be a good approach if you’re not sure how to combine the MESA8 indicators into profitable trading rules.

Neural network predictions learn patterns in historical data and are able to apply that knowledge to project price movement (or any other data stream used to train the net).  This model learns the patterns in the MESA8 Crossover Momentum, BandPass and Detrend indicators as they relate to the percent change in open 1 trading day into the future from the next open.  As in the other examples, the Crossover Momentum and Detrend indicators are more attuned to trending markets.

One other indicator is included in the prediction:  we used the range between the Crossover Plus Amplitude and Crossover Minus Amplitude in the belief that it would provide more information than the indicators by themselves.  

The training period (white line) for the model covered 6/2/03 – 10/31/05 and produced a 31.8% return on trades compared to 11.4% for buy and hold.   The out-of-sample period (green line) ran from 11/1/05 to 5/15/06 and generated an 18.7 % return on trades compared to 9.8% for buy and hold.