Saturday, May 27, 2017

The Market Is Not Broken

This post is my attempt to make sense of the interesting observation that many smart and experienced traders lapse into periods of trading like idiot rookies.  I don't think it's simply that their emotions get away from them or that they stop following sound processes.  In fact, I think it's just the opposite:  they keep doing what has worked in the past, but now--in changed market conditions--their strategies no longer produce an edge.  In other words, as market regimes change, consistency shifts from being a trading virtue to becoming a significant vulnerability.

Let's take a simple example.  I have created a daily measure of buying pressure and selling pressure from intraday uptick and downtick data.  I treat the upticks and downticks as separate variables reflecting buying and selling activity throughout each day.  My data set goes back to 2014 and we can examine how buying and selling pressure are related to price change X days forward.  Indeed, we can place buying and selling pressure readings into a multiple regression formula and identify an equation that significantly predicts forward price movement.

When we examine scatter plots of buying and selling pressure versus forward price change, however, we see significant departures from the linear regression line toward the extremes of the distributions.  In other words, when buying and selling pressure are unusually high or low, the implications for forward price movement are different than when the values are more moderate.  Methods that extend linear regression to identify significant nonlinearities are able to more precisely model the relationships among buying/selling pressure and future price change.  As it turns out, an important mediating (interacting) variable is the volatility of the market.  The relationship between past buying and selling pressure and forward price change is not the same in one volatility regime as in another.

So, for example, low volatility regimes see considerable momentum effects:  high buying pressure and low selling pressure tend to be associated with further price strength.  In higher volatility regimes, short term buying pressure or low selling pressure tend to be associated with short-term mean reversion.  In low volatility regimes, the most powerful predictive time horizon is between 10 and 20 trading sessions out--significantly longer than in higher vol regimes.

The point here is that the patterns we observe in markets do not have universal validity.  Whether we follow chart patterns and "setups" or statistical relationships, the predictive power of these varies as a function of market conditions.  When we move from a higher volatility regime to a lower one, for example, what used to work no longer has a universal edge.  The entire idea of finding your edge and trading it with flawless discipline and consistency is itself flawed.  We need to adapt to market conditions and find relationships specific to the conditions in which we find ourselves.

Lately I've heard many traders lament that the market is broken, that volatility is gone for good, that algorithms are manipulating prices, etc.  Meanwhile, they continue to apply their linear methods to a nonlinear world.  The stock market is not broken.  It is simply behaving like low volatility markets behave.  Edges are present.  They may not be the edges that were present several years ago, and they may not be edges on the time frame that you happen to prefer.  They also may not be edges that you can uncover with lines and patterns on charts or simple correlations and linear regressions. Our challenge is to adapt to what is, not stay mired in what used to be.