Trend Following…C’est la vie!
This summary is written by Rich based on a conversation in our CTA series between Philip Seager, Head of Absolute Return at Capital Fund Management (CFM), and the podcast hosts, Niels and Alan.
About Capital Fund Management
Niels and Alan were joined by Philip Seager, Head of Absolute Return at Capital Fund Management (CFM). They discussed CFM's background, including its start in 1991 as a trend follower and its evolution into a multi-strategy, multi-asset firm with $9.5 billion under management.
Philip shared that about $1 billion of their managed assets are in trend following products, and they have around 60 researchers and 120 IT personnel.
CFM is mainly based in Paris but is looking to expand its research staff in New York.
Phil discussed the firm's philosophy and their belief in trend following as a good strategy. CFM believes that inefficiencies exist in the market, but they are small and difficult to exploit.
They utilize a scientifically rigorous investment process, employing PhD scientists to tackle complex problems.
Philip explained that trend following has shown consistent performance over two centuries across various asset classes, and that periods of underperformance are statistically consistent with the strategy's modest risk-adjusted returns.
He suggested that a Sharpe ratio of 0.5 is reasonable for trend following, which is actually higher than the equity risk premium.
Philip acknowledged that the two centuries of trend following study may not be entirely realistic but still provides evidence for the existence of price return autocorrelations in assets over a long period.
"Two centuries of trend following" refers to a research paper that analyses the performance of the trend following investment strategy over a period of 200 years. The study investigates the historical performance of trend following across various asset classes, such as equity indices, fixed income, commodities, and foreign exchange (FX), demonstrating its consistent results and durability over a long period of time. The paper aims to provide evidence that the trend following strategy has been effective in the past and can still be relevant in today's market environment.
In the discussion, Philip talked about trend following's role in institutional portfolios and why it remains a smaller piece of their business despite its benefits. He mentioned that trend following struggles to gain a larger role in traditional portfolios due to its positively skewed profile, which can be challenging for performance chasers.
Philip believes trend following should be a buy-and-hold strategy and, if he had to choose one strategy to manage his own money, he would pick trend following due to its diversification and uncorrelated nature.
When they relaunched their trend following strategy in 2013, their added value was their experience in avoiding in-sample overfitting and their expertise in portfolio construction and execution. They aimed to use their implementation skills to build a better trend following product.
The Importance of Portfolio Sharpe
Niels brought up Cliff Asness' paper, which discusses concerns about the trend following industry focusing too much on Sharpe ratios and potentially modifying strategies for business reasons rather than performance.
Philip shared his thoughts on convexity in trend following, stating that it's a mechanical effect and that large infrequent gains and small frequent losses make the payoff resemble that of an option.
Philip mentioned that their firm has developed products for clients who want to maximize convexity with a small reduction in Sharpe ratio, and he believes that their offerings depend on the client's needs.
If clients want trend following for performance or as a hedge, they have different products to cater to those requirements. Philip emphasized that managed futures should not be restricted to trend following alone and that his firm uses various strategies to serve clients' diverse needs.
The Challenges of Overfitting in Research
Philip discussed the challenges of in-sample bias and overfitting in the context of trend following. He emphasized the importance of out-of-sample tests and having a solid understanding of the model.
In-sample bias occurs when a model is overly optimized to fit historical data, which may lead to poor performance in real-world trading. Overfitting happens when a model is too complex or tailored to the past data, making it sensitive to noise and less robust in dealing with new, unseen data. Both of these issues can result in a model that performs well in backtesting but fails to deliver the expected returns in live trading.
To mitigate these problems, Philip emphasized the importance of out-of-sample tests. Out-of-sample testing involves evaluating a model's performance on a separate dataset that was not used during the development and optimization process. This allows researchers to gain a more realistic assessment of how the model is likely to perform in actual trading conditions. By validating the model on unseen data, researchers can better determine if their strategy is genuinely effective or if it is merely overfitting to past data.
Additionally, Philip underscored the importance of having a solid understanding of the model being used. This means that researchers should be aware of the underlying principles, assumptions, and limitations of the model. By understanding the model's mechanics, researchers can have more confidence in its ability to capture and exploit certain market inefficiencies or behavioural biases. This understanding can also help them identify potential improvements or refinements to the model, ensuring that the strategy remains robust and effective over time.
Philip noted that in shorter-term strategies, higher levels of statistical significance and machine learning can be used without relying on an understanding of the model. However, for slower strategies, understanding the driving factors is crucial.
Regarding enhancing trend programs, he explaind that research is more focused on their flagship program, with development mainly occurring in areas like execution and portfolio construction.
Machine learning has provided value, but it has not been revolutionary. While machine learning has been around for a long time, its growing impact can be attributed to advancements in technology, computing power, and the availability of skilled researchers. Machine learning is particularly useful in execution, where shorter-term predictors are sought after.
The Factors that Create Differences in Trend Following Returns
Niels and Philip discussed the factors that create differences in returns among trend followers, specifically in the context of 2022.
Philip identified timescale, the portfolio of instruments, and the focus on cost and liquidity as key differentiators. He also mentioned that when volatility goes up, dispersion among CTAs tends to increase, albeit slightly.
Philip believes that trading in alternative or illiquid markets can generate performance, but if the flow of assets under management (AUM) into those markets were to stop, it could lead to a significant decay of performance. He highlighted that diversifying into more esoteric portfolios can result in less equity downside protection.
Niels and Philip also talked about the importance of risk control, focusing on stable risk and maximizing the Sharpe ratio while controlling for tail events. Philip explained that although their risk modelling is algorithmic and backward-looking, they also discuss potential tail events and market changes in a committee. If necessary, they intervene in their trading based on these discussions.
Role of Trend Following in a Multi-Asset Portfolio
Alan and Philip discussed the role of trend and managed futures in a multi-asset portfolio, examining if it's better to have trend exposure with capped equity or not. They find that incorporating a measure that takes drawdowns into account, such as the Sterling ratio, is helpful.
Their analysis suggests that equity downside protection is better in the mix when using metrics that don't incorporate drawdowns.
The Sterling ratio is a risk-adjusted performance measure that takes into account the average return and the downside risk of an investment. It is calculated by dividing the average annualized return of an investment by its average drawdown over a certain period of time. The higher the Sterling ratio, the better the risk-adjusted performance of the investment. The Sterling ratio is particularly useful for evaluating strategies that aim to limit drawdowns, such as trend following.
Philip acknowledged that non-equity markets can be important sources of the convex return profile in managed futures and trend following. He explained that trending on commodities provides a hedge against inflation and offers convexity against inflation.
When discussing the potential solutions for avoiding concurrent drawdowns in trend and equities, Philip admited it is difficult to determine the best timescale without the benefit of hindsight.
Capacity Limitations and Replication
Niels asked about capacity among managers and strategies and the potential pitfalls of doing replication on managers when only looking at some kind of regression of returns.
Philip mentioned that clients will rebalance their equity portfolio, but when trend following underperforms, they never reinvest in trend following, which is a sign that there’s not really traction in institutional portfolios for trend following.
On the subject of capacity, Philip said that the capacity of trend following is higher than most strategies as it's a directional strategy and trades the most liquid instruments in the world.
He mentioned that the risk of investing in a replicator is that the investor may not be investing with someone who has the skill set to deliver performance longer-term.
Misconceptions about Trend Following
Philip mentioned that firstly, it's not an easy strategy, and a reasonable degree of risk-adjusted returns is not easy to achieve.
Secondly, he disagreed with the idea of naivety on the convexity of trend following, saying that it is a statistical protection and not the same as options that give guaranteed payout during crashes.
The protection from trend following is for path-dependent events, especially long and protracted drawdowns, and crises that tend to play out on long timescales.
Lastly, he emphasized the importance of implementation skills in trend following. Philip emphasized that implementation skills play a crucial role in the success of trend following strategies. While a naive P&L may seem simple, generating risk-adjusted returns requires a significant amount of skill in implementation. He mentioned that there are numerous factors to consider when implementing a trend following strategy, such as trading costs, slippage, and other market frictions, which can significantly impact the returns. Additionally, he suggested that having a control over the execution of the trades, often quite sizable trades, is crucial for scaling up the strategy in the backtest and, therefore, crucial to be successful in implementation.
In other words, the devil is in the details, and it takes skill to implement the strategy effectively. In summary, Philip suggested that investors shouldn't underestimate the importance of implementation skills, and the success of a trend following strategy depends on the ability to execute it properly.
This is based on an episode of Top Traders Unplugged, a bi-weekly podcast with the most interesting and experienced investors, economists, traders and thought leaders in the world. Sign up to our Newsletter or Subscribe on your preferred podcast platform so that you don’t miss out on future episodes.
Most Comprehensive Guide to the Best Investment Books of All Time
Most Comprehensive Guide to the Best Investment Books of All Time
Get the most comprehensive guide to over 300 of the BEST investment books, with insights, and learn from some of the wisest and most accomplished investors in the world. A collection of MUST READ books carefully selected for you. Get it now absolutely FREE!Get Your FREE Guide HERE!