# How adaptive models got AlphaSimplex through the Covid crisis

## System sped up moves out of stocks into commodities and bonds

For trend followers like Kathryn Kaminski the defining feature of the coronavirus selloff was its speed – a feature that could easily trip a quant firm like hers. US equities plunged nearly 34% in 23 days. In the subprime mortgage crisis in 2008, the market took more than four times as long to fall roughly half as far.

“It was one of the fastest crisis periods we’ve had in the past 20 years,” the chief research strategist and portfolio manager at quant manager AlphaSimplex tells Risk.net. “As you can imagine, if you have a simple system that’s built to be the perfect system for 2008 or 2000, those parameters aren’t necessarily the optimal parameters in a very different environment.”

Trend-following strategies generate the signals they trade by comparing prices with long-term averages or to the past trading range of an asset. Quant firms must define the time periods over which reference points are calculated. Faster models look at more recent periods and react more quickly to market shifts. Slower models trade less during benign periods and so incur fewer transaction costs.

During a selloff like in March, adaptive trend following – an approach AlphaSimplex has pioneered – comes into its own, says Kaminski, who joined the firm in 2018 from the MIT Laboratory for Financial Engineering and serves as co-portfolio manager for the AlphaSimplex Managed Futures Strategy. Adaptive trend following seeks to remedy the problem of static trend following by allowing the system to speed up or slow down in changing environments.

“One of the challenges with the pure trend-following system is we often start with some pre-fixed time horizons and windows and ways that you define a trend. And if you have only those classic approaches, you’re going to pick strategies and approaches that worked well in the past but may not be the best solution in the future,” Kaminski says.

AlphaSimplex's managed futures strategy is up over 3% year to date, according to data from Bloomberg. Societe Generale's trend index is down 0.85%.

Based on AlphaSimplex's own analysis, a fast model went into the crisis with about a third of the exposure to the S&P 500 compared with a slow model (see figure 1), Kaminski says, returning nearly 8% compared with a 2% loss (see table A). That’s the biggest difference due to the speed of trend-following models of any crisis period the firm looked at.

In the first weeks of the year, AlphaSimplex’s system already detected shifting trends in the dollar, commodities and bonds. In late January it picked up on a small retraction in equities. Then, when the crisis hit, two things happened.

“First volatility spiked, which caused our systems to become less convicted to equities, causing us to take down our equity exposure through the end of February. In addition, our signals recognised a change in direction in the markets which [meant the system] reduced risk and pulled us out of equities,” Kaminski says. Returns ended up flat for the month of February.

Losses in equities and the process of getting out of positions were offset by gains from big moves in bonds and commodities. “We ended up relatively hedged in terms of P&L, which was a positive situation compared to many people in that scenario,” Kaminski says.

By March, AlphaSimplex had withdrawn almost completely from equities and was focused on opportunities in bonds and commodities. The adaptive approach meant the firm had changed the speed at which it traded. The system sped up the reduction of some of the positions in equities and moved faster into bonds as they continued to rally aggressively, Kaminski says.

Now the system has slowed again. The past two months have been characterised by a “wait and see” approach, Kaminski says.

“Obviously, equities have recovered quite a bit, but we’ve seen range-bound behaviour in the other asset classes. We haven’t seen the portfolio changing a lot. It feels like markets are waiting to see what happens through all of the stimulus, even though the equity market seems happily complacent again.”

AlphaSimplex’s trend-following strategy, which was launched in 2010, has grown to be one of the largest portfolios in the SocGen Trend Index with just under $5 billion in assets under management. Many traditional trend followers have made systems more complicated and added non-trend strategies over the past 10 years. But Kaminski’s research leads her to believe that pure trend tends to outperform in a crisis environment. The system is built to adapt but some changes are just too abrupt, Kaminski concedes. Kaminski says the short energy trade this year was an example. Oil started the year at$75 a barrel before oil futures plunged to below zero in April. AlphaSimplex’s system remains cautiously pessimistic on oil because the overall trend has been somewhat negative.

Did the system adapt fast enough to benefit from the recovery in stocks? “The signals are very confused about equities right now,” Kaminski says. “We have a much lower allocation.”

Even so, Kaminski trusts the adaptive approach is more likely to succeed through the current uncertainty.

“We tend to like periods of stress or dislocation,” she says. It is hard to predict how the ramifications of the coronavirus crisis will materialise. “If that’s the case, a trend-following strategy is going to do a good job.”

In periods of stress, volatility rises but that can create new trends and extend current ones, she says. Many strategies have a hard time navigating changing fundamentals and in moments such as this, trends in market prices can give an indication of where the market is moving. “The question we are asking ourselves now is whether this crisis is really over or whether there will be new trends that emerge.”

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