We were thrilled to hear that the NeuroShell Trader again won the Technical Analysis of Stocks and Commodities magazine's 2014 Readers' Choice award for best artificial intelligence trading software.  We couldn't have done it without you so we offer our sincere thanks for your support. 

I also want to share my appreciation to the developers and tech support staff for the many years they have devoted to making the NeuroShell Trader award winning software.  Most of the time it's a lot of exacting work interspersed with a flash of insight.  I don't believe there's another team quite like them anywhere else in the world and I'm very glad they are working on behalf of Ward Systems Group. 

My thanks would not be complete if I didn't include the wonderful staff at Technical Analysis of Stocks and Commodities magazine for producing an industry leading publication each month.  They have created a forum for discussion and education which has made our company successful. 

The Markets Are Changing

The steady upward progress of the markets that characterized 2013 has been replaced with a lot more volatility and we expect that trend to continue for a while.   Once again it's time to build trading models with NeuroShell Trader or ChaosHunter that can weather different market conditions. 

Customers who use real-time datafeeds such as DTN IQFeed and IBFeed and working with daily bars sometimes tell us that they can't decide on their trades for the next day's markets because the data vendor doesn't send a close bar until late in the evening.   (The eSignal interface is not affected since it is hard-coded and is capable of processing the end-of-session event.)  As you know, NeuroShell Trader issues a trading signal after the close of a bar.  But here's a tip that will let you get to bed before midnight and still let you send your trade to your broker in the evening.

Approximately 5 minutes before the end of market hours, save the chart with data.  (Go to the File Menu, select Save Chart As, and then be sure to check the box beside Save Data in the Chart.)  Close the chart, and then reopen it with the saved data.  The chart will use the last close saved in the chart as the day close and produce a trading signal.  Now you can just manually enter the trade with your broker.  You don't get the exact closing price used in the calculation of the trading signal, but you have the advantage of lining up your trades the night before rather than forgetting in the morning when the kids are late for school or bad weather makes your morning commute more hectic than usual. 

The next day, simply open the chart without the saved data to see how your model is working under current market conditions.

You can use your ChaosHunter models in NeuroShell Trader several different ways and the chart we are about to create will concentrate on two methods.  First we built a daily NeuroShell Trader for Deere & Company (DE), which displayed some regular price move over the selected period.  The regular price movement makes it easier for a neural network to find patterns compared to a stock that simply trends.

We built a neural network prediction of the open 1 period ahead using some of our favorite indicators from the Price Momentum Category (ADX, CCI, RSI, and Stochastic %K).  We set optimization to input selection.    The model found some trading signals at the peaks and valleys, including in the out-of-sample period.

Next we wanted to see which inputs ChaosHunter would choose.  We exported the price data from the NeuroShell Trader chart and added all of the indicators in the Price Momentum Category as inputs.   We selected all of the functions from the arithmetic category, negative and inverse from the algebra category, and two neural network functions. 

We let ChaosHunter run for a while and then created a new neural network prediction using the inputs found by ChaosHunter.  ChaosHunter chose Stochastic %K, MACD, and RMI.   That prediction is labeled Prediction ChaosHunter Inputs on the NeuroShell Trader chart that you may download from www.ward.net in either the New and Updated Examples section of the NeuroShell Trader or ChaosHunter section.  Those trading signals produced a decent profit in the out-of-sample period, but we wondered if perhaps we would have gotten better results if we had simply used the ChaosHunter Trading Signals based on the complete model rather than extracting the inputs from the ChaosHunter model and using them in a neural network. 

The Trading Strategy ChaosHunter at the bottom of the chart uses the exact ChaosHunter model by using the NeuroShell Trader's ability to call ChaosHunter models as indicators.  In NeuroShell Trader we selected the External Program & Library Calls indicator category and chose the ChaosHunter Trading Signal Indicator.  To use this indicator all you have to do is search for the ChaosHunter model, the .md file, which not only remembers the model but the prediction thresholds as well.  This model did not perform as well as the other two models in the out-of-sample period but did trade more often during the optimization period.

We leave it as an exercise for the reader to combine these three different trading models into a concensusTrading Strategy that says to buy when 2 out the 3 are true.  This method may possibly increase the chances of your overall Trading Strategy being correct.  Look for the Ensemble Estimators tip on either the NeuroShell Trader or ChaosHunter section of www.ward.net for an explanation of why this works.