From AI to cheese: funds seek fixes for trend following

Firms turn to machine learning, hybrid products and new markets to boost returns

  • The strategy has delivered positive returns in only three out of nine post-crisis years, leading some clients to question its effectiveness.

  • Trend followers are taking steps to boost returns, including investing in new markets, using machine learning and combining strategies.

  • Quant funds are also trying to reduce the implicit costs of market impact through smarter execution.

  • Many predict a growing bifurcation of trend following into cheap but simple and complex but more expensive products, with conventional programmes increasingly squeezed.

When silver futures plunged 10% in the space of minutes last year, they found a surprising big buyer: Transtrend, a systematic investment manager focused on trend following.

Surprising because trend followers usually operate over much longer timeframes – of weeks or months. Transtrend estimates it bought roughly 40% of what was sold on US venue Comex before the market bounced back.

In the past, the firm would have steered clear of such crashes or panics. Now it is “selectively active” when they strike, says its head of research Harold de Boer.

Transtrend is far from the only trend follower overhauling the strategy. 

“Telling clients you have boffins that piece through ancient scrolls of rice prices to fathom the enigma of trends doesn’t work any more,” says Matt Stevenson, head of product management at Florin Court Capital, a $650 million trend follower.

Having seen positive returns from trend following in only three out of the past nine years, some clients are questioning the approach – which, in broad terms, involves buying or selling assets based on movements in their medium- and long-term past average prices.

In response, quant funds are experimenting with ways to reinvigorate the strategy. They are pushing forward on four fronts: entering new markets, such as cryptocurrencies, ethane and freight futures; using machine learning to pick out hard-to-spot trends; boosting the strategy’s capacity to act as a crisis hedge; and improving execution.

At the same time, many firms are hedging their bets and also offer cheaper trend following programmes, whether revamped or conventional.

Pastures new

The first kind of innovation is the most common. Firms are looking to enter more exotic markets where many think trends are stronger and diversification easier to achieve. The greater the number of uncorrelated markets trend followers invest in – the lower their risk and the higher a programme’s Sharpe ratio.

Although not entirely new, the push into new markets has intensified in the last few years. Man AHL – the $24 billion systematic hedge fund arm of Man and one of the biggest trend followers – led the way with the launch of the first purely exotic fund in 2005. It unveiled another one in 2015. The newer fund trades assets that include cheese and the Kazakh currency.

In April last year, Florin decided to focus its portfolio only on exotic markets. The firm now trades trends in cryptocurrencies and in Chinese commodities via swaps with offshore banks. It is also eyeing freight futures and Indian commodities, says Stevenson.

More recently, in November 2017, Aspect Capital opened its trend following Alternative Markets Programme, ramping up to more than 180 markets since then. These include equity and fixed income exchange-traded funds, ethane and butane, which the firm started trading this year. The goal is to trade more than 200 markets, including some particularly unconventional ones, says Christopher Reeve, who leads the firm’s work on bespoke investment products.

GAM Investments and Systematica Investments have launched similar strategies.

Devet Capital, meanwhile, has expanded the notion of exotic assets, moving into “synthetic” trends in more conventional markets. The firm trades movements in spreads between different equity indexes, between equity and sector indexes and between one commodity and another.

“We believe there are still pockets of liquidity and potential that have not been fully explored: trading synthetic markets using mainstream futures,” says Leonardo Marroni, Devet Capital’s chief operating officer and chief risk officer.

Sharper and smarter

Trend followers are also trying to sharpen their strategies by finding new ways to spot trends. Some are using machine learning for this purpose.

Campbell & Company is one example. It has found machine learning helpful in determining how to group markets together to check for the existence of corroborating trends across different assets.

A trend with solid foundations should show up in signals across various assets, as a fundamental driver lifting one asset will usually lift at least one other asset as well. In one possible scenario, a period of favourable weather might boost a commodity, such as grain futures, which could in turn boost the local currency. But knowing where to look for such signals can be difficult.

“There may be no obvious reason why, on any one given day, markets like the S&P 500, gold and the South African rand should be trading together,” says Joe Kelly, who heads bespoke products at Campbell & Company.

At AlphaSimplex, roughly 40% of models use machine learning to group markets, identify trends and judge how to balance risk across them.

The approach notched up a success in May when the models worked out that Italian government debt had “decoupled” from other European bonds, says the quant fund’s chief research strategist and portfolio manager Kathryn Kaminski. Shortly afterwards, the Italian bonds sold off heavily, while their European peers remained supported.

 Kathryn_Kaminski.jpg
Kathryn Kaminski

This sensitivity to changes in market behaviour is befitting a fund founded by Andrew Lo, the author of the adaptive markets hypothesis. The hypothesis describes markets as a constantly evolving system, driven as much by human behavioural traits and biases as by certain more or less immutable laws.

Still other firms aim at improving the strategy’s ability to offset investors’ losses during different types of market downturns.

Trend following became famous as an effective hedge during the financial crisis, but has fared less well in recent, less prolonged and more isolated market dives, exacerbating rather than softening the impact of reversals. For example, the SG Trend Index plunged nearly 9% in February as previously upward-trending US stock markets made a U-turn.

The market flip would have been particularly painful for trend followers’ pension fund clients with big equity exposures.   

Trend followers are now looking to combine trend with other strategies such as market-neutral equity, for example, which could balance out its exposure in a stock-market run-up.

Some are also looking to tailor the strategy to a client’s individual offset needs. As Matthew Sargaison, chief executive of AHL, puts it, “one client’s crisis is not another’s”.

A new breed of tailored products could include an equity-neutral version of a trend strategy, or a capped quality strategy, or involve going long volatility less expensively than buying outright protection, he says. There is also the possibility of incorporating actively managed credit hedges to protect against risks in a client’s own industry, he adds.

Such products might not even be “fully systematic”, Sargaison notes.

Campbell & Company is taking a different route to the same goal: it offers a pure trend following programme designed explicitly to emphasise equity diversification.

The strategy is volatility-scaled based on equity risk, scaling up exposures in more volatile stock markets and down when volatility is low. The idea is to maximise the offsetting impact of the strategy when stocks are tumbling. This “dynamic risk targeting”, as the firm calls it, contrasts with the usual trend following approach of dialling down risk when markets are choppy.

Cost busters

But tackling the returns side of trend following is only half the story. Funds are addressing the cost side as well.

The focus here is on minimising market impact.

This can be significant. Winton, a firm that’s made its name in trend following, has decided to downscale the strategy in its main fund partly as a result of such implicit costs. In February, fellow trend followers exited positions en masse, exacerbating market turmoil – this became the “straw that broke the camel’s back” in driving the firm’s decision, according to Winton’s chief executive David Harding.

AHL is using machine learning algorithms to route orders in the most cost-efficient way possible. The algorithms’ benefit is their ability to adapt to a changing live environment.

Transtrend has overhauled its execution and now its models scan price data to systematically identify the trends the firm wants to be in, but also to select the best instruments and points in time to enter a particular market, taking into account market impact. The goal is to act as a liquidity provider, on balance over time, extracting more favourable pricing.

 Harold-de-Boer-large.jpg
Harold de Boer

“We don’t want to be the one paying the liquidity premium. We never have orders that say we will buy or sell irrespective of the price,” de Boer says. This sets the firm apart from less patient players, such as passive index tracker funds – since they are often required to rebalance daily, they will usually trade shortly before the market close regardless of price.

In some scenarios, Transtrend’s models could help it collect a particularly high liquidity premium – for example, if the firm fills the liquidity gap when others, such as high-frequency traders, take flight.

And at least one trend follower believes its management of market impact gives it the edge over competitors. Crabel Capital – a $2.8 billion hedge fund known for its short-term quant programmes – decided to add trend following to its strategies partly because it thought incumbents might be underestimating the market impact of their trading.

Crabel has identified over $300 billion worth of assets that are managed by trend followers and examined how they are deployed across timeframes, industries and trading venues, says Grant Jaffarian, a portfolio manager.

He declines to say what exactly Crabel does with this market map, but says the firm’s approach is more complex than simply executing individual trades so as to minimise market impact. “We are trying to extract the trend premium as cleanly as we possibly can,” he says.

For some, such steps to modernise trend following may still fall short of transforming the strategy’s performance. “It’s unlikely further research will magically ‘fix’ significant periods of underperformance,” says GSA chief executive David Khabie-Zeitoune. “It’s about marginal improvements rather than dramatic new insights.”

Alongside wider efforts to revitalise the strategy, many firms have launched lower-cost, often simpler versions of trend following. Some are doing only that.

Most recently, Winton has added a management fee-only fund to its trend offerings, following in the footsteps of Aspect, Campbell, CFM, GAM Investments and Systematica. Meanwhile, the trend following products of GSA Capital Partners, Crabel and AlphaSimplex are all fee-only. 

Fees can be unusually low: only 50 basis points of assets under management in some cases. 

Down but not out

Many expect products will increasingly gravitate towards the two current extremes: cheap and simple, on the one hand, and complex, higher-fee products on the other. In this view, the middle ground occupied by conventional programmes charging both a fee and a percentage of returns will continue to be hollowed out. 

“There remains demand for trend but at a much lower fee,” says Anthony Lawler, co-head of GAM’s systematic investment unit. On the other hand, there is also demand for “interesting things like esoteric markets”.

Others point out that just being cheap is unlikely to be enough. The quality of risk management and portfolio optimisation even in a simple product will mark out winners and losers, says Jean-Philippe Bouchaud, chairman and chief scientist at CFM. “If you take a formula from Wikipedia and trade with that, that’s not going to work.”

Old-fashioned versions of trend following in high-fee portfolios are becoming increasingly rare.

CFM’s hedge fund no longer trades anything recognisable as traditional trend. Winton is cutting in half the exposure in its main fund to trend following.

Smaller rival Devet Capital is choosing to trade a combination of trend following and mean reversion. “We are not saying traditional trend following doesn’t work and should be avoided like the plague. But we are saying it makes sense to have a smaller component of it than you would have done five or 10 years ago,” says Marroni.

Almost by definition, when people are ready to give up on it, trend following tends to have its best performance
Kathryn Kaminski, AlphaSimplex

Whatever the future structure of the trend following sector, it has reasons to hope for brighter times ahead.   

There are some fresh ideas on the drawing board, such as going long or short the very ‘trendiness’ of markets, based on estimates of when a trend – any trend – will set in or when a trend will be snuffed out by arbitrageurs.

“Trend doesn’t work all the time, so the idea is to identify robust indicators that allow you to trade trend as an asset,” says CFM’s Bouchaud, drawing a parallel with how managers today trade volatility as an asset.

Secondly, as quantitative easing is unwound in the United States and Europe, assets should stop moving quite so much in tandem, opening up more distinct trends for trading. Diversification should then be much easier to achieve. 

And lastly, given the right circumstances, trend following may yet surprise its detractors.  

The strategy comes into its own when markets undergo a deep-rooted reversal with wide knock-on effects, such as the one brought on by the oil price collapse in 2014, says Kaminski at AlphaSimplex. That is because the approach can trade the new trends, making its followers money while other investors are still trying to pick up the pieces.

With volatility low and US equity markets breaking record highs, the time may be ripe for just such a flip. 

Kaminski points out that in 2014 people were talking about how trend following was dead. In the end, it was a stellar year for trend followers.

“Almost by definition, when people are ready to give up on it, trend following tends to have its best performance.”

What went wrong?

Between 1987 and the crisis year 2008, trend followers recorded only two loss-making years, according to BarclayHedge’s systematic CTA index. Between 2009 and 2017, they’ve had just three positive years. At the same time, the strategy’s failure to offset recent stock market dives has got clients questioning its much-vaunted credentials as a source of ‘crisis alpha’.

What changed?

Not everyone ventures an answer. Those who do blame either overcrowding, quantitative easing, changes in information flows, a glut of liquidity, or electronic trading – or some combination of these.

Many think trends have become rarer – that coordinated actions by central banks to prop up economies after the crisis have caused markets to move in lock-step, melding trends together and robbing firms of diversification.

For instance, Transtrend’s Harold de Boer says the firm now trades as few as three trends across markets but the trends are more diverse and the firm’s exposures are bigger than before. “If you can find seven really different trends, it’s a lot,” he says.

Kathryn Kaminski at AlphaSimplex disagrees there are now substantially fewer trends than in the past, noting that signal-to-noise ratios show only a slight weakening since 2008. But trends are harder to discern because today’s markets tend to fixate on big stories that drown out weak signals from nascent moves. Information spreads faster now and “things are more contagious”, she says, citing social media as a possible reason. 

An alternative view is that abundant liquidity is smothering trends before they get going.

“There’s a huge battle going on to provide liquidity: through market-making, high-frequency trading, electronic markets. There are many more people trying to trade the bid-ask spread. It might be killing the trends in their nascent state,” says Jean-Philippe Bouchaud at CFM.

Yet another theory blames erratic markets for playing havoc with trend follower’s statistical models.

“Traditionally, the larger the price move – the higher the information level,” says Harold de Boer at Transtrend. “It’s almost the opposite way around now: often the larger moves don’t have lots of information content but no content at all,” he explains, referring to flash crashes.

One of the reasons prices have become more erratic and volatile is the demise of floor trading, de Boer says. In the old days, brokers would speak to each other and their clients in the morning about what they expected to trade that day, creating some common expectations for the trading session. 

“That’s gone and it has a big impact on the opening of all kinds of markets. Look at the US stock market. It’s no exception for the daily range to be set in the first five minutes.”

Some in the sector, however, suggest a simpler explanation for trend followers’ troubles in the past decade: it is the nature of the beast. 

“If you plot a strategy like this in a simulation spanning several decades, you’ll find it has fallow years,” says David Khabie-Zeitoune at GSA. “It’s just the way it is for a low-Sharpe ratio strategy.”

Editing by Olesya Dmitracova

  • LinkedIn  
  • Save this article
  • Print this page  

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact [email protected] or view our subscription options here: http://subscriptions.risk.net/subscribe

You are currently unable to copy this content. Please contact [email protected] to find out more.

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here: