Coursera is one of my favorite sites online. Two weeks ago, I finally had the opportunity to register for the course that I’ve been waiting for: Digital Signal Processing.
Digital Signal Processing is used in everything from music to communications to… predicting financial markets. Although it’s not the focus of the class, I already drummed up a few ideas worth exploring as indicators.
The first indicator idea that the course inspired, and the one that I’m the most excited about, looks for bars that are out of place with their peers. I arbitrarily chose a lookback period of 20 for my analysis.
Steps to calculate the cumulant ratio:
- Calculate the average for periods 1-20
- Calculate the sum of the absolute value of Close of the bar – average for bars 1-20
- Calculate the sum of the Close of bar 0 (the last closed bar) – close of each bar, 1-20
- Divide step 3 by step 2
The initial result looked as expected. I placed the absolute value bars on step 2 so that I wouldn’t have to worry about the average changing the sign of the output. The top function is only positive if the current price is “above” the sum of the last 20 bars. A negative position means that the current price is “below”.
Most markets are noise, which creates a natural noise band between ±1. Prices way out of line with the average show huge jumps like the one in the image.
Values around ±1 are expected. You wouldn’t, after all, expect a new bar to throw off the calculations very much.
That’s exactly the point. If something is near the ±1 window, then it’s probably worth ignoring. The price action is pure chop.
The real value is in the spikes. A spike of 20 isn’t any more important than a spike to 50. Above a certain threshold, the trader just needs to pay attention. It’s a black and white issue.
Strategy Ideas
My initial instinct was that the indicator would work well on emerging market currencies like TRY and ZAR. They’re so breakout prone that entering on the right side of the move, regardless of the reason, should do well.
It took about 20 minutes of chart gazing to come up with the strategy for the H4 chart. It was flat on most currencies, but USDZAR stood out. This equity curve used my initial settings without optimization. More importantly, the profit factor improved on a 1.5 year walk forward test.
The breakout strategy did great on USDZAR H4 charts
The majors showed exactly the opposite. They do trend, but hardly in the mega-monster manner of the emerging markets. I modified the strategy to do the opposite and let ranging conditions prevail. EURUSD, as seems to be usual with range trading strategies, stood out as the best performer.
The range trading approach works best on EURUSD H4
Nerd section
For the math nerds and programmers out there, the formula for the indicator is below.
The equation for the cumulant ratio