Thursday, October 04, 2018

Where Does Market Edge Come From?

Above is a screenshot from my trading station, capturing roughly the last eight days of trading in the ES futures, using hourly data.  Volume is on the X axis.  Below volume is a two-period RSI that is a very short term measure of overbought/oversold.  Above volume is a linear regression line (red) with bands of one standard deviation (blue) above and below.

The chart nice illustrates how markets are a joint function of linear trend (the regression line), with one or more cyclical functions overlaid (the movements from upper to lower bands; the movements between short-term overbought and oversold).  Notice how often the market will take out a prior low or a previous high, only to cycle back the other way.  It's a good illustration of how traders who think primarily in terms of breakouts, momentum, and directional trends get "chopped up" by the market's cyclicality.

Much of the frustration from trading comes from the imposition of linear thinking on markets that have significant cyclical drivers.  During stable market periods, we can "solve" for dominant cycles and then use movement above and below the bands to identify when stability is breaking down and we're entering a different cyclical or directional regime.  The idea is *not* to trade according to your personality.  The idea is to capture the linear and cyclical components of markets and trade the market's personality.

If you get the basic idea of interacting trends and cycles, you can see why the following has very limited edge:

*  Chart patterns taken out of context;
*  Oscillator readings taken out of context;
*  News "catalysts" taken out of context;
*  Breakouts taken out of context;
*  Wave counts or Fibonacci patterns imposed on markets;
*  Releases of fundamental data

So what does have edge?  In my experience, it comes from identifying when 1) the herd (which thinks one-dimensionally in terms of linear/directional movement) gets trapped at cycle extremes and 2) when shorter-term "mean reversion" traders get trapped by the momentum components of longer-term cycles and trends.

What has edge in the quant world is the dynamic modeling and re-modeling of shifting cyclical and directional components of markets, rather than static, curve-fit models based on lookback periods.  This is a big reason why funds that make use of true machine learning are the top performers in recent years and why many funds underperform their benchmarks.

Further Reading: