Use the prediction parameters dates tab to set up the data ranges for the prediction paper trading and optimization.

The parameters on this tab may or may not be visible depending upon the currently selected Prediction Wizard Interface Options. Press the Options button to change the Prediction Wizard Interface Options. For more information see Prediction Wizard Interface Options.

1. Specify the Data used to evaluate prediction.

You can specify that the data range(s) match the chart, be a specified size, or start and stop on specified dates.

2. Specify whether or not to Save optimization which performs best on later paper trading.

If you choose Save optimization which performs best on later paper trading, the model’s parameters are still optimized on the optimization set, but each new optimal solution that is found by the GA is applied to the paper trading set. If that optimal solution is found to get better results on the paper trading set than previous optimal solutions, then it is saved as the ‘best model’. Optimal solutions that underperformed on the paper trading are still used in the GA optimization process to find an optimal solution on the optimal data set, but they are not used as the ‘best model’. The final model selected by the optimization is the last saved ‘best model’.

3. Specify whether or not to Start trading before last chart date.

This option may or may not be visible depending upon the currently selected NeuroShell Trader program options. To enable this option, from the main NeuroShell Trader menu, select Tools, choose Options, navigate to the Advanced option tab and check Allow real trading to begin before last chart date under the Date interface settings.

The Start trading before last chart date option allows a ‘trading’ statistics data set to be calculated after the optimization and paper trading data sets. This ‘trading’ data set will be a true “out-of-sample” data set. It can be used if you do not wish to keep the model that did best on paper trading, but yet you still want to see how a model did on data other than the optimization data set. This data set can also be used in conjunction with saving the model that did best on subsequent paper trading to evaluate the model and yet another independent data set.

4. (DayTrader Only) Specify whether or not to Restrict training and signals to a specified time range.

When you select the Restrict training and signals to a specified time range option, NeuroShell will ignore trading signals outside of the specified time range and will automatically send an order to close out any open positions just AFTER the completion of the first bar whose time stamp is equal to or greater than the specified end time. Note that an overnight time range can be specified by setting the start time after the end time (for instance start at 4:00pm and end at 9:00am).

Note that if you wish to have a position fully closed out before a certain time, you will need to specify an end time earlier than that time. As an example, if you have 1 minute bars from 9:30 to 4:00 pm each day, if you specify 4:00 as the ending time, a closeout trading signal will be generated on completion of the 4:00 bar, which will result in the actual trade fill not occurring until the next bar at 9:30 the next day. To ensure the fill occurs before 4:00pm, you would need to set the end time to the bar timestamp prior to the 4:00 bar, which in this 1-minute example would be 3:59.

Additionally, selecting Restrict training and signals to a specified time range limits the training data presented to the prediction to the specified time range. However it should be noted that even though the training data is limited to the time range, the underlying inputs and indicators are still calculated on the entire data range. If you wish to limit the time range on which the inputs are calculated, then you must limit the time range of the underlying chart data (see Format Chart Dates )

When you are satisfied with the prediction parameters, press the OK button to return to the Prediction Wizard.

Note that the optimization range of data is important because it determines which trading strategy is the “best” (as specified by your Optimization objective on the Optimization tab). The default is to optimize across the same data as the paper trading data, as this seems to be the standard among our competitors. This is a little risky, because we are really looking at in sample results. On the other hand, the NeuroShell Trader Professional is, we believe, the first software to optimize trading strategies that actually offers an alternative. In other words, we are the first to provide a paper trading data set that is out of sample from the optimization data set. Therefore, you might want to take some risk here since it has been standard practice for some time. If you do not want to take that risk, simply do not choose the Save optimization which performs best on later paper trading option.

Notes:

• It should be understood that trading can result in losses, and that past performance of a trading system is no guarantee of future performance.
• When selecting a given number of days, it should be noted that the days are measured in days of data and not calendar days. So if you specify 7 days, and the market you monitor does not have data over the weekends, you will get 7 full days worth of data and not just data from a Monday to Monday.

 
Topics of Interest:
What are Neural Networks?
Neural Network – Network Criteria Discussion (General)
Neural Network – Network Criteria Discussion (Objective)
Neural Network – Network Criteria Discussion (Advanced)
Troubleshooting Your Model – What to Do if You Feel You Haven’t Been Successful