Is There Any Future in Equity Index Trend-Following?

The problem of trend following is fundamentally simple: choppy markets reduce the effectiveness of trend-following algorithms and shorter trend durations reduce their profitability. There is no need for a more sophisticated analysis. About 15 years ago I derived an equation that describes the problems of trend-following. In a nutshell, since the duration and magnitude of future trends cannot be known in advance, trend-following methodologies with higher win rates are required for survival.

Basic math

By design, most trend-following methods have low win rate. There is a clear trade-off between the win rate w and the payoff ratio R. As one attempts to increase the win rate, usually the payoff ratio drops (the ratio of average win to average loss). As a result, the profit factor PF, which is given by the following formula:

PF = w × R / (1 – w)   (1)

either remains the same or even decreases. From equation (1) it may be seen that if the win rate is 0.35, then to get profit factor PF = 2, the required value of R is

R = (1-w) × PF / w   (2)

or after plugging in the numbers, R = 3.175

But if w is increased to 0.50 then as a result R must be at least equal to 2 to maintain the same profit factor of 2.  Anyone who develops trading systems knows that achieving such performance is very difficult although not impossible. In my opinion, 99.99% of trading system developers will never achieve performance defined by {w=0.5, R=2}. Some can get as high as {w=0.5, R=1.50}. At that level, the difference between robustness and noise trading is not exactly distinguishable in the short term (meaning with a couple of years of trading data).

An example

An example of a trend following system that shows how recent conditions have affected the robustness and efficiency of such methods is the 50-200 MA cross system. Basically, this is a long/short symmetric system that establishes long positions when the 50-day MA crosses above the 200-day MA and exits and reverses when the opposite happens. This system has worked very well in SPY since 1994 and even during 2008 with a return close to 34%. But after 2009 this system stopped performing well because, as will shall see via the help of a special indicator I have developed, the market dynamics changed drastically for a period of almost two year. During that period the system was subjected to a large drawdown of about -38%.

To make a long story short I have placed everything on one chart below:

SPY_20130402_MA50_200

I summarize the main points of the chart:

(1) This simple system worked very well up and until 04/2010
(2) The bottom pane shows the downtrend in the duration between two consecutive MA crosses of the system and the downtrend is clear.
(3) The two small shaded circles show periods of prolonged choppy action.
(4) The situation seems to have improved a bit with the last cross having a duration of 292 days while it has moved above the critical level of 200.

Thus, from the chart it is clear that market dynamics have changed after 2010 and the particular system has been unable to adjust. Despite that the overall performance figures are still good with a CAR of 8.31% for a fully invested system in SPY ETF (Number of shares is based on 100% of available closed equity). Note that after 01/2010 the win rate drops to 50% but the payoff ratio decreases to 0.57, resulting in PF <1, a plain disaster. Below is a table of the system monthly returns:

SPY_20130402_MA50_200_MO_Returns

Causes

There are mainly two causes of the problems trend-followers have faced in the last few years in equity indices but also in commodity markets:

(1) HFT robots arbitrage out serial autocorrelation reducing trend duration
(2) Financial instability increases choppiness

Cause (1) is here to stay because HFT is the modern market maker. Cause (2) will come and go but never disappear.

Summary

The indicator I used in this article that measures the duration between MA crosses can be adapted to most trend-following methods to analyze performance in the context of market dynamics. Trend-followers should be alert as the times of easy money are over and market dynamics have changed substantially with more uncertainty and less autocorrelation. There is some hope on the horizon for this year but this may change at any time.

Disclosure: no relevant position at the time of this post and no plans to initiate any positions within the next 72 hours.. 

Charting program: Amibroker (Charts created with AmiBroker – advanced charting and technical analysis software. http://www.amibroker.com/”) 

Disclaimer

Related posts

http://www.priceactionlab.com/Blog/2012/12/the-main-cause-of-failure-of-some-popular-technical-trading-methods/
http://www.priceactionlab.com/Blog/2011/01/why-trend-following-is-hard-a-quantitative-answer/

This entry was posted in Trading Strategies, Trend following and tagged , , , , , . Bookmark the permalink.