Investing in Hedge Funds: Is the trend your friend?
Since its 2009 lows, the S&P 500 has achieved a total realised gain of over 175%. That is in the 98th percentile for all five-year rolling periods since its inception in 1928.
It seems only natural to assume that trend-following systems should have capitalised on this surge. Has this been the case? The answer rests on the speed of the trend-following system applied to this market.
This analysis will allow us to gain insight into performance differentials across various trend-following systems, with a focus on trading speeds.
Additionally, we evaluate the long-term performance of trend-following systems and demonstrate the benefits of speed diversification within a portfolio.
Fast, medium or slow?
To perform this analysis we utilise three different trend-following systems – fast, medium and slow, applied solely to the S&P500. Figure 1 examines the Sharpe ratios and correlations with the S&P 500 for the different trading systems, both during the recent five-year period and since the inception of the index.
While the slow system has performed better over the past five years, this system embodies a buy-and-hold strategy, as signified by the high correlation to the underlying market. The recent outperformance of the slow system is at odds with the long-term Sharpe ratios, which are better for the fast and medium systems. To review this outperformance from a historical perspective, figure 2 displays the outperformance of various speeds through time, on one and five-year rolling Sharpe ratios.
These suggest that it is difficult to predict when a system will provide future outperformance. Throughout the course of history, including the most recent five-year period, slow systems have outperformed the least often, with either fast or medium-term systems outperforming slow systems 73% of the time, on five-year Sharpe ratios.
To put the past five years into perspective we can look at the risk-adjusted return distributions for each of the speeds. Figure 3 plots a histogram of the five-year Sharpe ratios for fast, medium, and slow systems since 1928. For each histogram, the recent five-year period is marked in grey for comparison. The histograms show that the slow system has recently outperformed its historical average while both fast and medium systems underperformed.
To delve deeper into the properties of the systems we can examine the return distributions. The distribution for fast systems’ returns is different from tthe other models. Faster systems do not exhibit normal distribution. They are positively skewed with lower dispersion. Medium-term systems exhibit almost no skew and slow systems exhibit a negative skew on a scale similar to that of the index itself – which is intuitive, considering that as holding periods become longer they more closely resemble a buy-and-hold strategy.
Figure 3 also demonstrates that there is substantial dispersion in performance of trend-following systems across trading speeds throughout history, implying a degree of diversification that can be achieved by deploying different trading speeds. An equal-weighted combination of the three systems achieves a 25% improvement in risk-adjusted returns, with a Sharpe ratio of 0.58.
Does adding trend-following to a long-only portfolio improve performance? To answer this question we look at the historical performance of the three systems and our equal-weighted ‘Combo’, all respectively mixed in a 50/50 split with the S&P 500 (figure 4). Holding the S&P 500 since inception in 1928 has had a Sharpe ratio of 0.37 (as depicted by the black line).
Unsurprisingly, the addition of trend-following to the long-only portfolio causes improved risk-adjusted performance, and diversifying across systems and the index provides the best risk-adjusted returns. The combination of the S&P 500 with the fast system comes a close second. It becomes evident that the portfolio benefits of trend-following are diminished as the system becomes slower.
It is undeniable that the previous five years have been a joyous ride for the long-only equity investor. A review of trend-following performance across speeds demonstrates that slow systems outperformed during this period – but not necessarily over a longer time period. Meanwhile, fast and medium-term systems would have found this trend difficult to follow, but have performed better over the entire period.
It remains impossible to predict when one system will better another, but it is clear that as systems become slower, the portfolio benefits will diminish: slow systems tend to have a higher correlation to the market itself, providing marginal diversification benefit to an equities investor.
The skews show that the slow system’s returns, very much in correlation with the S&P 500 itself, will be subject to large unexpected drawdowns as the system takes too long to react. Alternatively, the fast and medium systems will be in a much better position to profit from crisis – this is shown by the 2007-09 global financial crisis in figure 2.
Given the difficulty of predicting the future, diversifying across a wide spectrum of time frames provides a solution to improving long-term performance of trend-following. Moreover, we find the addition of a trend-following allocation significantly increases the risk-adjusted returns for a long-only investor.
Alex Greyserman is chief scientist at ISAM, adjunct professor of mathematics at Columbia University. He is also the co-author with Kathryn Kaminski of Trend following with Managed Futures: The Search for Crisis Alpha (Wiley 2014)