Paid to be Paranoid: Trending, Fast and Slow

Higher volatility levels represents a real challenge for trend managers. Here’s how allocators can assess some of the different types of trend models managers use.

Introduction

Over what timeframe should you determine a trend? When considering active trend following managers, we believe it is beneficial for allocators to investigate how exactly each strategy has been designed and in what market conditions it might best perform. Even in the same markets and over the same period, differences in the methodologies could result in significantly varied investment results.

Let’s take for example the rollercoaster year of the 10-year US Treasury yield and what a ‘successful’ following of the trend might have looked like:

Figure 1. Following the 10-Year US Treasury Yield (%)

Problems loading this infographic? - Please click here

Source: Bloomberg and CitiVelocity; as of 29 October 2021.

To have successfully followed this trend, a manager would have been short treasury yields up to the end of March, long until early August and the short again for the remainder of the year. Not an easy task.

How to Make Trends and Influence People

Whether a trend manager is successful in navigating different markets often depends on the speed of their trend calculation. Methodologies that account for a longer time window tend to outperform in slowly trending markets. However, when volatility is elevated, having a ‘faster’ model that is more responsive to shifts in market directions can be better.

In our hypothetical trend-following example, imagine we have three distinct strategies to compare, each tracking the 10-year US Treasury yield.

Figure 2. What Would These Models Have Recommended?

Problems loading this infographic? - Please click here

Source: Bloomberg and CitiVelocity; as of 29 October 2021. Note: Hypothetical Results are calculated in hindsight, invariably show positive rates of return, and are subject to various modelling assumptions, statistical variances and interpretational differences. No representation is made as to the reasonableness or accuracy of the calculations. Since trades have not actually been executed, Hypothetical Results may have under- or over-compensated for the impact, if any, of certain market factors.

Figure 2 illustrates that the ‘Slow’ model started the year with a strong signal for investors to go long on US 10-year futures, while ‘Medium’ and ‘Fast’ were a lot less sure that this was the right call. When all the models pointed investors to go short over the summer, only Fast changed direction briefly before returning to short.

Wake Me Up When September Trends

Turning the signals of these models into investments gives us a chance to test them. In other words, what would have been the result if a portfolio had followed these signals to the letter? By using the signal strength as a proxy for the level of exposure and assuming a monthly rebalance for each model, we can see which trend model has worked the best so far this year (Figure 3).

Figure 3. Back-testing the Trend Models (Cumulative Profit and Loss)

Problems loading this infographic? - Please click here

Source: Bloomberg and CitiVelocity; as of 29 October 2021. Note: Hypothetical Results are calculated in hindsight, invariably show positive rates of return, and are subject to various modelling assumptions, statistical variances and interpretational differences. No representation is made as to the reasonableness or accuracy of the calculations. Since trades have not actually been executed, Hypothetical Results may have under- or over-compensated for the impact, if any, of certain market factors.

Since the Slow model started the year with an especially strong buy signal at a time when Treasury yields were rising, this model would have lost value and only continued to worsen. Conversely, the Fast and Medium models made quick gains – especially since they were short in February and March – however, the ability to react quickly to the mid-year change in direction resulted in Fast maintaining a positive result year-to-date.

I Get by With a Little Help From my Trends

It would be fair to say that these hypothetical trend models are simplistic. Bestpractice trend following strategies make use of other techniques to help account for the unpredictable nature of markets playing havoc with their models. A common feature is volatility scaling: when markets are calm, the model increase position sizing to maximise the return potential from smaller movements. Conversely, when markets are highly volatile, the model reduces the size of the positions.

We can overlay a simple volatility scale onto the Fast, Medium, and Slow models to see whether this would have helped or hindered the US 10-year strategy.

Figure 4. Applying Volatility Scaling to the Trend Models

Problems loading this infographic? - Please click here

Source: Bloomberg and CitiVelocity; as of 29 October 2021. Note: Hypothetical Results are calculated in hindsight, invariably show positive rates of return, and are subject to various modelling assumptions, statistical variances and interpretational differences. No representation is made as to the reasonableness or accuracy of the calculations. Since trades have not actually been executed, Hypothetical Results may have under- or over-compensated for the impact, if any, of certain market factors.

The application of volatility scaling helped the Fast model deliver an additional 6.5% cumulative return. The Medium return was slightly less negative as a result, but the Slow model was instead worsened even further by the volatility scaling, losing more than 87% of invested capital since the start of the year.

Does this mean that allocators should always favour models with fast decays? Not necessarily. For example, consider a sharp equity sell-off that rebounds to its previous position in the space of a week. Models with faster decay periods would have cut exposure and therefore missed the subsequent reversion while slower models remained invested and benefited. Scenario planning and consideration of the broader market context – such as levels of volatility – are key when considering which trend-following models might perform best.

Conclusion

The lesson for allocators in this hypothetical exercise is that higher levels of volatility represent a real challenge for trend managers. In our view, it is helpful to have trend models that can adapt their speeds for the changing market environment and be able to ‘lean in’ to positions when things are moving in the right direction. Being able to dive into the details of each manager’s strategy to understand the environments which might prove more beneficial or difficult for returns should be a crucial part of an allocator’s risk analysis.

Ultimately, there is no one-size-fits-all approach to trend following. We believe that allocators should be clear-eyed about which models will most align with their investment beliefs.