I am working on a collection of other indicators that use the ramp-weighting of WMA, such that the oldest bar in the lookback interval gets the least weight, and the newest bar gets the most. I intend to post this collection in the file-sharing area, but I need suggestions on what to include.
Exponential weighting (like EMA), I've noticed, takes too long to settle sometimes, particularly when adapting it to statistical measures like standard deviation.
Simple or constant weighting (like SMA) is terrible. It generates false signals. A constant weight indicator will signal when a large value enters the lookback interval, and then it signals again when that value leaves. This is especially true for statistical indicators.
So far, I have written and tested the following 'weighted' counterparts to the traditional constant-weight indicators:
- WATR - weighted average true range
- WBollinger - weighted Bollinger bands
- WKeltnerChannel - weighted Keltner channel
- WLinReg - weighted linear regression
- WLinRegSlope - weighted linear regression slope
- WStdDev - weighted standard deviation
- WStdError - weighted standard error indicator
The above are, for the most part, non-trivial. The trivial ones are where you simply replace any occurrence of EMA or SMA in the code with WMA (WATR ad WKeltnerChannel are the only 'trivial' ones above). CCI would be trivial. But are there any other non-trivial ones I can implement?
-Alex