• Trend-following can be used as an objective methodology to seek to participate with market growth and protect against significant declines
  • Common models may work on average but can fail in very specific scenarios
  • Parabolic moves are particularly tough for simple trend following methods to capture
  • We believe a dynamic approach to trend following is critical for long-term success

While all eyes were on the Federal Reserve this week, we took notice of another interesting development.

The 50-day moving average finally crossed below the 200-day moving average for the Dow Jones China Broad Market Index.

Despite peaking in June and crumbling nearly -45% since, it was not until this weak that a long-term sell signal was given in this common trend following model.

So what happened?

From a very high mathematical perspective, trend followers tend to think of stock prices as being some part trend and some part noise. By using something like a simple moving average, a trend follower can average out all the noise to identify the underlying trend.

This practice will generally work if (1) noise isn’t so large it swamps out the trend, and (2) the trend is linear.

What do we mean by linear? Consider a car traveling a constant 60mph on the highway. Given its current position and speed, we can estimate how far down the road the car will be a minute in the future: 1 mile. Now, maintaining a constant 60mph is difficult, so the car may be a little over or a little under 1 mile.

We can capture this with a simple mathematical model: distance traveled = 60mph * time traveled.

But what if the driver has their foot to the floor and the car is accelerating? Our simple model becomes an awful representation of reality because the speed of the car is no longer constant. The distance traveled becomes parabolic and so we need to use a parabolic equation to capture the dynamics of the process.

What does this have to do with trend following?

Simply put: China went parabolic.

2015-09-19 - Figure 1

And a parabolic move is tough for a simple trend following process because it is similar to trying to track the location of the car without taking into account the acceleration. Consider how far behind the 200-day moving average fell:

2015-09-19 - Figure 2

One simple solution is to just use a shorter moving average. By using less historical data, the shorter moving average can keep up better than a longer version (though is arguably therefore more sensitive to whipsaw risk).

2015-09-19 - Figure 3

This is why many trend followers use a “cross-over” technique. They look at both a long and a short moving average and go long when the short moving average is above the longer one, and sell-out (or even short) in the reverse scenario.

The problem is that when the price dynamics go highly parabolic, even this method fails. At peak, the price was 58% higher than the 200 day moving average. Quite simply, the 200 day moving average no longer represented the long-term trend.

This is why we believe so strongly that a dynamic method is required for tracking trends.

In Newfound’s momentum model, we balance the strength of the trend we’re tracking against how much noise we think there is in the price process.

When the trend is very strong and noise is very weak, our models “shorten up” and try to track price more closely.

When the trend is very weak and noise is very strong, our models “loosen up” and try to incorporate more data so the model can average out the noise and seek to avoid whipsaw.

The result is similar to a moving average that can at times behave like a 200 day moving average and at times behave like a 50 day moving average depending on the price process.

Consider the following image, which shows the underlying price estimate dynamics for Newfound’s trend model. You can see that during the parabolic price increase, the model looks much more similar to the 50 day moving average we saw above.

2015-09-19 - Figure 4

In fact, we can even plot how much data our model is using at any given time.

2015-09-19 - Figure 5

(Left axis is ETF price while the right axis is the number of days our model is using to calibrate).

We can see that as China’s market price went parabolic, Newfound’s model shortened the amount of data it utilizes, changing its behavior from being similar to a 300 day moving average to a 100 day moving average.

While parabolic markets can be difficult for most trend followers to navigate, taking a dynamic approach gives us a fighting chance.


Corey is co-founder and Chief Investment Officer of Newfound Research, a quantitative asset manager offering a suite of separately managed accounts and mutual funds. At Newfound, Corey is responsible for portfolio management, investment research, strategy development, and communication of the firm's views to clients.

Prior to offering asset management services, Newfound licensed research from the quantitative investment models developed by Corey. At peak, this research helped steer the tactical allocation decisions for upwards of $10bn.

Corey is a frequent speaker on industry panels and contributes to, ETF Trends, and’s Great Speculations blog. He was named a 2014 ETF All Star by

Corey holds a Master of Science in Computational Finance from Carnegie Mellon University and a Bachelor of Science in Computer Science, cum laude, from Cornell University.

You can connect with Corey on LinkedIn or Twitter.