Where are the Trends?
In his interesting research paper “Where are the Trends?” (attached - click "Image Not Available" box above) Martin Estlander seeks an answer as to why his trend program is in the worst drawdown in its history. Many traditional trend followers are in a similar situation.
His answer is that the past two years have seen very few strong trends in the aggregate – at least as regards:
- His program’s portfolio; and
- The 6 month trends which he seeks to capture.
He sets out to show, graphically, the historic frequency of strong 6 month trends since his program’s inception in 1992 using the following methodology:
“Let us take a look at how the opportunities for trend following have been. There are many ways of measuring market trendiness. Here we choose a trend opportunity measure that counts the number of strong trends based on the return over the past six months from the highest high or lowest low to the current price, for each of the instruments. We then add relevant instruments for each market segment. A higher reading means that a larger number of meaningful trends have occurred and a lower reading means a lower number of aggregate number of meaningful trades over the past six months.”
I am not at all certain from this description how “strong trends” were measured for these purposes and so I decided to take a slightly different approach, which seems to have produced very similar results. I took a reasonably balanced portfolio of around 100 futures, some of them going back as far as 1970. I used ratio adjusted prices. For each day of the test, instruments which were at a 6 month highest high or a 6 month lowest low were assigned a score of “1”; else “0”. For each day of the test, such scores were added together and the sum was divided by the number of instruments for which data existed on the relevant day. Thus expressing the number of instruments at 6 month highs or lows as a percentage of the total portfolio of instruments as it stood on the relevant day.
I then smoothed the data in three different ways to produce 3 different but very similar charts. I attach the spreadsheet I used and display the graphs below.
“Daily Average” uses a 255 day moving average of the daily % score and a polynomial regression trend line based on the daily data.
“Monthly Average” groups the daily % score into an average for each complete month of the test and then shows a 15 period moving average of the monthly results, together with a polynomial trendline.
“Annual Average” groups the daily data into an average % score for each complete year of the test and simply plots each year’s average on the chart. It also features a polynomial regression line to emphasize the trend.
For clarity, linear charts have been used which exaggerate the trends but the results clearly support Martin Estlander’s findings that the large number of strong trends peaked in 2008 (on an annual basis) and hit a low in 2012.
This piece of research forms part of a series I have been producing to look at why the past two years have been so poor for traditional long term trend followers.
Further thoughts on the matter can be seen at the following URLs:
I also intend to follow this post with an analysis of the serial correlation of price movement in futures instruments since 1970.
The overall conclusion I have reached is that long term trend following has undoubtedly become more difficult over the past 40 years. I won’t speculate further on the well rehearsed and frequently aired reasons for this, nor on the fact that the past two years seem to stand out as a particularly difficult period.
Some analysts believe that trend following is a dangerous illusion doomed to failure and that trends simply do not exist in a purely random market; or in an “efficient” market.
My own belief is that markets are neither perfectly efficient nor wholly random in the longer term and that trends will continue to unfold with changing economic circumstances and evolving valuation assumptions. I believe that more frequent stronger and longer trends will once again emerge and that traditional long term trend following will resume its profitability; but methods will need carful adjustment over time to take account of changing market conditions.
As I have often freely admitted, I am neither a mathematician nor trained statistician – merely an ignorant market participant and observer. I welcome any suggestions to correct or improve my analysis and would like to hear the reasoned views of others on this topic.