Technical Indicators in Other Trading Platforms

Top  Previous  Next

ChaosHunter lets you control how certain technical indicators behave when they cannot compute due to the division by zero condition when used in other programs or trading platforms.

 

The six affected indicators are: ROC, ROC%, %Chg, Spread%, Relative Strength, and Efficiency. When ChaosHunter is installed initially, it will always output "*" (missing value) when any formula that contains these indicators encounter a division by zero condition. However, you can modify ChaosHunter so that the formulas it creates can be compatible with TradeStation (and other trading platforms) where there is no so-called "empty/missing constant".

A program parameter (boolean flag) that regulates technical indicator behavior is called ZeroIndicatorMissingValue and is available through the ChaosHunter.ini file, which is included in the directory where you installed ChaosHunter. Users can change value of the ZeroIndicatorMissingValue flag by manually editing the ChaosHunter.ini file in a text editor.

The ZeroIndicatorMissingValue parameter recognizes two values: True and False.

ZeroIndicatorMissingValue=False (default) instructs the program to output "*" (missing value) when encountering division by zero.

ZeroIndicatorMissingValue=True forces ChaosHunter's built-in indicators to output the number zero when encountering division by zero.

When saving the model on the disk, the ZeroIndicatorMissingValue flag is saved along with it. So, each model knows what was the value of the ZeroIndicatorMissingValue flag during optimization. As a consequence, when loading a model from the disk and firing it, the model's flag takes superiority over the flag in the ChaosHunter.ini file.

The runtime server applies the model's ZeroIndicatorMissingValue flag appropriately. This ensures compatibility between ChaosHunter and external programs (such as NeuroShell Trader) that employ the runtime server to fire ChaosHunter formulas.

Please note that the ZeroIndicatorMissingValue flag has no effect when indicators cannot compute due to insufficient data at the very beginning of the time series or when the input time series has non-numerical "*" in them. ChaosHunter still outputs "*" in those cases.