The Real Secret behind Trend Following and How it Works | with Martin Lueck

ELI5/TLDR

Martin Lueck co-founded AHL (now part of Man Group) and later Aspect Capital — two of the longest-running systematic trend-following shops in the world. In this conversation with Niels Kaastrup-Larsen, he argues that the real edge in trend following isn’t the entry signal, it’s the holistic plumbing: position sizing scaled to volatility, execution that hides you in the market, portfolio construction across 150-plus instruments, and a risk team whose job is to shred every new model the researchers propose. The strategy works because it’s boring and repeatable, not because anyone is clever.

The Full Story

The position is the stop

Most retail conceptions of trend following imagine a signal that says BUY, an entry, a stop-loss a few percent below, and a ride that ends with either a profit or a stopped-out loss. Lueck spends a good chunk of the interview gently dismantling this picture. Aspect doesn’t use stop-losses in the conventional sense. The position itself is the stop.

What sits underneath the position is a continuous function of two things: signal strength (how confirmed the trend looks across multiple time scales) and volatility (how much risk a unit of position represents right now). If volatility doubles, Aspect cuts the position in half to keep the risk constant. If the trend weakens, the position bleeds smaller. Nothing snaps — everything modulates.

“It’s a gradually modulating signal… knowing effectively when to back off and perhaps when to be a provider of liquidity to the markets rather than a consumer of liquidity — that’s another delicate feature of what we do.”

The second half of that quote matters more than the first. At Aspect’s size, you cannot just slam orders into the market when your signal fires. You become the market. Which leads to the second secret.

Execution is half the game

Aspect runs roughly 96% of its trading box-to-box — algorithmic execution monitored by a human trading desk 24 hours a day. But the goal of those algorithms isn’t speed. It’s invisibility. Lueck calls it “capturing a patient premium.” The execution algos dribble trades into the market in random pieces at random delays, deliberately trying to look like noise rather than a billion-dollar CTA rebalancing.

The reason becomes obvious once you hear it. In the early AHL days, bank trading desks figured out their models and front-ran them. Aspect realised the same effects could be captured with a much smoother entry and exit, effectively invisible. Four years of research effort (as of 2017) went into the transition to fully algorithmic execution. They’re not competing with HFT shops on milliseconds — they swap algos out on the order of months, not minutes — but they care a great deal about the average daily volume they’re absorbing and the spread they’re paying.

150 markets, seven sectors, no view

Aspect trades over 150 markets — closer to 200 instruments when you count multiple contracts per market — across seven internal sectors. The default portfolio is aggressively agnostic. Lueck’s starting point is an even spread across sectors. He then “flexes very gently” around liquidity constraints and long-term correlation structure, but he refuses to overweight bonds just because bonds have been working.

“We absolutely defy or eschew the worldview that says you can look at what’s happened historically and predict what’s going to happen in the near future… your ability to infer that ‘well, because I’ve had a good run in bonds means they’re going to continue to perform, so I’m going to go overweight in bonds’ — that’s not something we ascribe to.”

There’s a small irony there which Lueck doesn’t dwell on: trend following is itself a bet that recent behaviour persists. But the bet is at the level of individual market moves, not at the level of sector allocation. The portfolio is built to capture trends wherever they appear, not to predict where they’ll appear.

Eighty-twenty: trend plus modulators

Of the risk budget, 80% goes to medium-term trend-following models. The remaining 20% sits in what Lueck calls “modulating factors” or “complementary models.” These aren’t faster trend models or counter-trend overlays designed to get out earlier. They’re orthogonal factors with similar persistence — relative momentum between instruments, carry effects in FX and fixed income, features of the term structure, occasional mean-reversion at faster timescales. Some are correlated with trends, some aren’t. The point is to capture diversifying premia that have their own reason to exist, not to nudge the trend models around.

This is a careful distinction. Many CTAs run “trend plus some other stuff” and the other stuff turns out to be discretionary risk on/off overlays. Aspect is trying to stay systematic across the whole stack.

Drawdowns and the customer’s emotional life

A long stretch of the interview is about drawdowns — both how they happen and how investors react to them. Lueck tells a story about Peter Lynch’s Magellan Fund: the fund itself had a legendary track record, but a study of actual investor flows found that retail money had been roughly net flat, because everyone bought after the leaps and sold after the dips.

“If you aggregated those chunks of money that typically bought at the highs and sold at the lows… it was nowhere like the performance that the chap had generated.”

His framing for clients is that managed futures earns its place in a portfolio because it has low correlation to equities and bonds. When equities crash and trend following makes money, you’re a hero. When equities boom and trend following bleeds, you’re an idiot. Both are the same product. The asymmetric reaction is the price of admission for the diversification.

This is fund-marketing language, and Lueck is honest about it being fund-marketing language. He’s making the case that an investor who bought managed futures for the diversification should accept the drawdowns; an investor who bought because it was the only thing that worked in 2008 will sell at exactly the wrong moment.

The risk team’s job is to break things

Aspect’s internal structure has a research team that builds new models and a risk team whose job, Lueck says, is to shred those models. Each new component — a tweak to a trend signal, a new modulator, a portfolio construction change, an execution algo update — goes through formal review with out-of-sample data, synthetic data, failure-case examination. Three releases a year aggregate the approved components into a new program version, and the risk team then reviews the holistic release: how everything works together.

He calls it an academic peer-review atmosphere. The risk team members are paid to be as smart as the researchers and to assume the researchers are wrong.

The deeper test is historical. Could you have run this version of the program through 1994, 2005, 2009? If the new model would have concentrated risk into a sector you couldn’t have held through the drawdown — if a client would have called and asked what the hell you were doing — then you can’t launch it. You have to add a constraint or modify the mix.

Drawdowns inside the manager’s own head

Niels asks a question that gets a more candid answer than usual: is there anything that keeps Lueck up at night? The official answer is operational — exchange outages, geopolitical events, the limits of fiduciary care. The unofficial answer is more interesting. Lueck talks about complacency as the underrated risk. You should always expect to see something new. The psychological challenge isn’t just the client’s — it’s the manager’s, and the discipline is to not start believing your own models when they’ve worked for a stretch.

A small confession about discretion

One of the more useful exchanges, late in the interview, is Lueck’s take on the “I’m 95% systematic with a discretionary overlay” claim that some CTAs make. He’s blunt about it.

“If somebody says to you ‘I’m 95% of the time systematic and I’ve got a discretionary overlay’ — no, you are 100% discretionary, and you will be unable to get the confidence about the integrity of the models that you developed and the usefulness of any back-tests that you do.”

The argument is structural. If you reserve the right to step in when things go wrong, you can never back-test honestly, because the back-test assumes the model runs unattended. Once a human can intervene, the strategy is the human, not the model.

Key Takeaways

Aspect Capital’s flagship is roughly 80% medium-term trend following plus 20% diversifying “modulators” (relative momentum, carry, term structure features, occasional mean reversion). The modulators are not trend-tweakers — they capture orthogonal premia.
Position size is a continuous function of signal strength divided by volatility. When volatility doubles, position size halves. There is no discrete stop-loss; the position itself bleeds smaller as conditions change.
Execution is treated as a research problem on par with signal generation. Box-to-box algos dribble orders in random sizes at random delays, deliberately obfuscating Aspect’s footprint. The goal is to be a patient liquidity provider, not a fast consumer.
150+ markets, ~200 instruments, seven internal sectors, agnostic default allocation. Sector weights flex only for liquidity constraints and long-term correlation structure, not for recent performance.
Three program releases per year. Each component is reviewed in isolation; each release is then reviewed holistically by a dedicated risk team that is structurally adversarial to the research team.
Historical replay as a gating test. Could you have run this version of the program through 2008, 1994? If a client would have pulled the plug on the drawdown, the model can’t ship.
The structural case for trend following is behavioural persistence. Markets trend because participants under-react and then over-react to information; this persists across decades because it’s wired into human decision-making, not because anyone has found a clever pattern.
“95% systematic with a discretionary overlay” means 100% discretionary. If you can intervene, you cannot back-test honestly.
Investor flows destroy investor returns. The Peter Lynch dollar-weighted-return story applies to CTAs too. The drawdown discipline is the actual product.

Claude’s Take

The CTA marketing is loud in places — the “democratization of managed futures,” the casual mention that Aspect is “well placed” for whatever comes — but Lueck is more candid than the genre usually allows. The discretionary-overlay takedown is worth the price of admission. So is the honest framing that trend following’s payoff is the price of accepting the wrong-time drawdowns, not some clever pattern recognition.

The conversation is most interesting when Lueck is technical: position sizing as a continuous function rather than a stop, execution as a deliberately slow patient process, the 80/20 split with modulators as truly orthogonal factors rather than trend-tweakers. These are not secrets — they’re industry-standard at the top tier — but hearing them stated plainly by someone who has been running them since the late 1980s carries weight.

Where the talk gets thinner: the “intuitive” defence of medium-term trend following (“if you read the FT, you’ll roughly understand our positions”) feels like a soundbite for institutional consultants more than a real claim. A program that trades 200 instruments across seven sectors with vol-scaled positions is not intuitively legible, and saying it is mostly serves the marketing brief. The drawdown-emotional-management section is genuine but standard for the genre — every CTA gives a version of this talk.

The most interesting piece for someone outside the systematic trading industry is the organisational design. The research team builds, the risk team breaks, three releases a year, every component reviewed in isolation, every release reviewed holistically, every new version stress-tested against the worst historical episodes. The discipline isn’t about the alpha. It’s about not blowing up while collecting the alpha.

Score: 7/10. A solid interview with a thoughtful operator, useful as a window into how a serious systematic shop actually runs, but the format (long podcast, host gently leading) means the texture is industry-conventional rather than revelatory. Worth listening to if you want the institutional CTA worldview straight from the source; not essential if you already know the contours.