Adapting to evolving market regimes

Hamlin Lovell
Originally published on 07 September 2021
  • Above: Harold de Boer, Head of R&D and Managing Director, Transtrend

The market environment around and after the coronavirus crisis has presented challenges and opportunities, but operationally it has been business as usual at Transtrend. Remote working for most of the staff has not caused any disruption to investment or trading. Transtrend trades 24 hours a day and has always had some people working from home monitoring trades through the night, while others have the flexibility to work at home if they are carers or need home repairs. Remote working has however made creative thinking and integration of new team members somewhat more difficult: “Some of the best ideas can come from colleagues who themselves are unaware of the importance or application of their thought, but it’s recognized as valuable and shaped into a gem by others in the team. This sort of team dynamic is hard to replicate in virtual meetings,” says Transtrend Head of R&D and Managing Director, Harold de Boer. Nonetheless, Transtrend has engaged in some deep thinking about how to adapt its models, systems, asset class weightings and investment universe to changing market regimes. 

Maths is not the same as numeracy, because data measurement is subjective, and the usefulness depends on understanding the story behind the data.

Harold de Boer, Head of R&D and Managing Director, Transtrend

Prominent market themes in 2020 and 2021, such as inflation and energy transition, could mark regime changes that may wrong-foot investors. Their strategies can explicitly or implicitly be biased to the low inflation climate of recent decades. For instance, low inflation might not only bias asset allocation towards fixed income, but the associated low interest rates could also incline benchmark-conscious equity investors towards long duration growth sectors and stocks including many technology names. Relatively simple and static systematic models could be too rigid to adapt. More active and advanced systematic trading approaches should have the flexibility to adjust.

Maintaining exposure to capitalize on post-Covid trends

Once bitten by the Covid crash, some managers and systems, including some alternative risk premia (ARP) approaches, then shied away from risk-taking and underperformed. Others embraced the brave new world. “It has taken some effort to maintain enough exposure to take advantage of new trends. The Covid shock created potential to over-estimate forecasts of both volatility and correlations, which could have resulted in sub-optimal position sizing, and extremely low portfolio volatility. We then would have missed out on some big trends that were generated by the crisis and its policy responses. If one does not absorb risk as indicated by higher volatility, there is no risk premium,” says De Boer. 

Some traditional approaches to forecasting volatility and correlations, such as the EWMA (exponentially weighted moving average), which heavily weight the most recent observations, could have markedly downsized positions after March 2020 and resulted in too small positions. Also in 2021, short-term volatility, associated with phenomena such as cryptocurrencies or speculative retail investor frenzies in certain “meme stocks”, could have resulted in systematic strategies reducing or exiting exposures.  

Transtrend was not entirely immune to these typical portfolio management dynamics but anticipated the consequences of such periods of extreme turbulence. Initially, in March 2020, Transtrend’s undiversified risk (adding up position risks assuming 100% correlation) did drop as some positions were cut or exited, but by May 2020, it had increased to about 50% above pre-coronavirus crisis levels. The ability to not only rebuild exposure but also augment it distinguished Transtrend from some systematic strategies targeting constant forecast volatility that might well have undershot their target volatility in 2020.

Measuring volatility and other moments

Transtrend’s Covid response was partly informed by many years of experience, including research that showed how lowering the program’s volatility could, paradoxically, result in deeper drawdowns, partly because the systems do not have a chance to recover from setbacks. “We compare trading with the Dutch national sport of ice skating. Volatility measures how fast a skater goes, and a drawdown is falling behind in the race. Skaters need to go faster on slippery ice to catch up after they lag behind,” explains De Boer.

But volatility is too broad and generic a term to capture the nuances of how Transtrend measures its portfolio performance aspects from multiple angles. “Maths is not the same as numeracy, because data measurement is subjective, and the usefulness depends on understanding the story behind the data. We do not use standard deviation to estimate volatility, but rather prefer to use measures that avoid over-reacting to a short-term spike in volatility. Volatility measures have also been adapted in 2020 to leave more flexibility for trend indicators to work,” says De Boer.

In any case volatility is only the second moment of the distribution and only one lens for viewing risk. Skewness, the third moment, and kurtosis or “fat tails”, the fourth moment, are also important. After March 2020 the kurtosis in many markets came down while their volatility increased, which is the inverse of what happened in the central bank support driven market environment after the financial crisis.

Measuring correlation in a non-linear and non-Gaussian world

All of these and other metrics can feed into portfolio construction as can correlation. “Position sizing, which is explicitly or implicitly based on forecast correlation, is one of the most important considerations distinguishing different CTAs and their performance. Back in the early days of CTAs, simply calculating a correlation matrix might provide some competitive advantage. Now that everyone has access to computer power to do this, the question is more about the meaning behind the numbers. We do not use linear correlation measures such as Pearson correlation because markets do not behave in a linear way. Systems need to allow for asymmetric patterns of volatility and correlation,” says De Boer. For instance, many commodities have a positive skew and are more likely to “crash upwards”, but in fact Transtrend research shows that no markets move symmetrically in the manner of a normal distribution. 


Transtrend has traded what might be dubbed “alternative markets” since its Diversified Trend Program began in 1992 when even futures themselves or commodities in general were considered alternative.

There is no single silver bullet solution for the variety of distributions observed. “We use a variety of correlation measures, all of which have advantages and disadvantages, but the combination gives more chance of weathering a change in regime. I like to hear the research team saying that correlation measures are not completely under control – if they become confident that the systems are under control, they are more likely to run into trouble,” says De Boer.

Indeed, there is no certainty that historical correlation patterns will persist. Systems need to be flexible enough to cope not only with new correlation patterns being observed, but also those that have never been seen during back tests. “We will not rule out the possibility of equities and bonds declining together, even if we never saw this in a backtest. We are alert to the possibility that some parts of our program might implicitly assume this because it did not happen before. The same applies for inflation, energy transition or political change – where politicians may be gaining more power than central banks,” says de Boer. 

Investors should not place too much confidence in their correlation forecasts. Realised correlation after the March 2020 crisis proved to be much lower than a forecast extrapolating the most recent data. For example, Transtrend observed decent levels of dispersion within equity markets: between banks and healthcare, between airlines and Amazon, and within commodity markets: between crude oil and natural gas, and between wheat and palm oil.

Defining different trends and different markets 

These cases of divergence allowed Transtrend to identify some more isolated trends, but others were more difficult to separate from bigger themes. There are often overlaps between trends and themes so how broadly or narrowly to define trends and themes, and therefore whether positions are adding concentration risk, is a big challenge. This was especially the case after Covid overwhelmed markets, because the common factors became so large that it is more difficult to perceive depth and disentangle individual trends. 

“The objective is to maintain sizeable positions in different trends,” says De Boer, and the investment universe can help with this. Trading more granular markets allows for more precise exposure. In equity markets, Transtrend trades a mix of sectors and single stocks to pick up different trends. Transtrend has also for many years been conceiving some synthetic markets that are hybrids straddling asset classes, such as various short bonds, long commodities combinations that performed well in 2021.

It’s not about predicting inflation, but about being prepared.

Harold de Boer, Head of R&D and Managing Director, Transtrend

Adaptive strategies

Whichever markets are traded, any strategy needs to adapt to new regimes, in terms of the macroeconomic climate of inflation or industry dynamics of energy transition or in relation to the portfolio impact of new patterns of volatility and correlation. This challenge is common to both discretionary and systematic managers, as well as those using elements of discretion to drive their systematic process. “Whether programs automatically adapt, or humans intervene to adjust models to adapt, is fundamentally not different. Ultimately in both cases it’s humans that provide the required adaptivity,” says De Boer. 

Systematic strategies in particular could now have become less adaptive because they are overly automated. Static rules are not well equipped to navigate new regimes, and hypothetical backtests covering multiple decades or even centuries may reveal an inflexible mindset wedded to rigid systems. “Studies showing centuries of evidence in favour of trend following using models that don’t need to be adapted repeatedly marked the peak of hype around the strategy. In reality the world is constantly changing, and nobody can in fact backtest new phenomena such as cryptocurrencies, populist politics, energy transition, or the Covid pandemic. In fact, such fundamental changes are ultimately driving trends, now and in the past. On top of that simulations tend not to correctly consider crucial variables such as market impact or transaction costs. It has always been easy to make simulations look good and computer power now makes that even easier. What has really changed and what is truly concerning is the academic credibility given to simulations these days,” says De Boer.

Transtrend’s active and adaptive approach was able to make substantial shifts during the second quarter of 2020 based on price trends in individual markets and bigger picture themes. “We switched from predominantly risk off in April 2020 to mainly risk on by June 2020 and this was a gradual process. Iron ore started showing an uptrend which later broadened out into an Asia Pacific economic recovery theme driven by Asian and Australasian economies that contained Covid by closing their borders. Separately the European and US recovery theme was more driven by government support through fiscal and monetary policy. After November 9th 2020 vaccine approval, this morphed into a recovery and reopening trend, which Transtrend profited from partly through exposure to equity sectors and single stocks,” says De Boer.

“We like to think in terms of robustness rather than speed.”

Inflation and commodities

Inflation illustrates how Transtrend approaches a regime change. It does not monitor aggregate market inflation as an input since its models are based on the technical price history of individual markets, not on fundamental economic data. Therefore, the debate around whether various official measures of inflation might be understating true inflation is somewhat academic. And Transtrend does not trade anything explicitly based on inflation measures, such as inflation linked government bonds or inflation swaps, but has profited from the pickup in inflation through other asset classes. “It’s not about predicting inflation, but about being prepared,” says De Boer.

Commodities are clearly one obvious arena for inflation and have been Transtrend’s largest asset class exposure since the second quarter of 2020. Some systematic funds have predetermined limits on asset class exposures which place a cap on commodities, but Transtrend is more opportunistic in expanding the allocation according to the strength of signals and thinking more about trends and factors. If commodities are the clearest way to gain exposure to a theme such as recovery, reflation or inflation, then they should be sized larger than other asset classes. 

Iron ore and Malaysian palm oil were amongst the biggest winners in 2020. Transtrend also maintained substantial positions in smaller and more peripheral commodity markets such as lean hogs and cattle and French rapeseed oil. “These markets seem small now but back in the 1970s they were some of the largest futures markets,” says De Boer. Transtrend is not trading local Chinese commodities but trades many other markets that are driven by the nowadays huge global impact of developments in Chinese supply and demand. 

Though broad-based commodity exposure could have worked well in late 2020 and early 2021, Transtrend does not view commodities as a homogeneous asset class, and they are certainly not solely based on the inflation/reflation theme. “Weather factors such as droughts in 2021 have been important for markets such as corn, coffee and sugar. We do not look at commodities as one asset class. We do not trade commodity indexes because everything in the index can be traded individually and price discovery in some smaller markets can be disrupted by large index flows. We do not wish to contribute to that. Most of the time crop futures like in wheat offer a large risk premium on the short side due to the term structure,” says De Boer. 

Alternative markets sit with traditional markets

Some commodities are sometimes classified as “alternative markets”. Transtrend has traded what might be dubbed “alternative markets” since its Diversified Trend Program began in 1992 when even futures themselves or commodities in general were considered alternative. Within the commodity complex, Transtrend has for some years been trading carbon emissions, Nordic electricity and German electricity, which some might label as “alternative markets”, just like iron ore or palm oil. Transtrend however groups all markets together in its program and argues that it is not helpful or useful to separate some markets and define them as alternatives. There is a danger that the definition of alternative markets is driven by the desire to market an exotic or unusual strategy, rather than a sensible and logical way to define them.

De Boer points out that there is no consistent way to define alternative markets and at least eight criteria could be used, which could relate to the size or type of markets or the venues on which they are traded: smaller market size; non-Western exchanges; unusual underlying markets; newly accessible markets; non-standard contract specifications; synthetic markets such as spreads between markets; off exchange liquidity pools and OTC markets. 

If each of these criteria is followed through to the logical conclusion to generalize the definition, it then becomes easy to identify inconsistencies in the criteria used to distinguish alternative from conventional markets. For instance, most developed market currencies are traded OTC but are not described as alternative. Some markets such as orange juice are small and fairly illiquid but not dubbed alternative. Sometimes the same commodity now trades on multiple western and non-western exchanges. Futures exchanges have always been launching new contracts, which have included Treasury bonds or equity indices that are now viewed as very conventional. 

Transtrend’s investment universe is designed to efficiently capture different underlying trends and not based on whether markets are defined as traditional or alternative. It might add or delete both traditional and alternative markets. In recent years Transtrend has somewhat streamlined its investment universe to prioritise differentiated markets and has ceased trading some alternative markets. Equally it does trade some smaller markets such as sunflower seed futures that have marginal impact on performance, naturally paying careful attention to liquidity since smaller markets can be more vulnerable to disruption caused by flows. 

Transtrend strives for a balanced portfolio. In doing so, it questions whether “alternative markets” are necessarily or always less correlated with traditional markets. They may on average have lower correlations but can become correlated during key turning points such as those seen in 2020 around February 23, September 2, and November 9. This suggests that there are in fact some common factors driving returns for these markets. 

When markets are aggregated into larger factor exposures, it seems beneficial to trade “alternative” and “traditional” markets together within a more diversified program, rather than in standalone programs. For instance, though iron ore, which might be labelled as “alternative”, turned out to be the biggest winner in 2020 it is also possible that a more conventional market, the New Zealand dollar, might have been the largest winner in that year. Both uptrends were arguably primarily a function of the Asia Pacific recovery trend. Transtrend would rather keep an open mind about which markets will show the best trends than restrict a program to a subset of markets or carve out a subset of its flagship strategy to offer separately.

There have been many changes over the past 18 months but not everything needs to change. There are arguments about whether short-, medium- or long-term trend followers will perform better in different climates. Transtrend has not felt the need to change the speed of mainly medium-term models. Adaptiveness should not be conflated with the concept of speed. Intuitively, shorter term systems might sound more responsive, but they will not necessarily pick up new trends, partly because they can be more prone to being distracted by noise. Transtrend has also not extended into longer term time frames because research showed they would not perform better. “We like to think in terms of robustness rather than speed, we want to ignore short term noise and stay in long if the opportunity is there, but we want to be agile and turn quickly when we need to as well. You don’t solve that with the concept of speed, that’s too one-dimensional and there are so many different dimensions to modeling,” de Boer concludes. 

With this open minded and fluid approach to systematic investment management Transtrend continues to adapt to new market regimes.