Simplicity in Trend Following | Andreas Clenow

ELI5/TLDR

Clenow’s biggest regret about his first book is that the trading model wasn’t simple enough. The cleanest trend-following rule in the world is two data points: is today’s price higher than the price one year ago? If yes, go long. If no, go short. Run that on fifty futures markets with volatility-based position sizing and you will beat most of the elaborately optimized systems on the market. The whole interview is a sustained argument that complexity is something you have to be paid for, and most traders are not getting paid.

The Full Story

The complexity tax

Clenow opens with a thesis he comes back to over and over: complexity is a cost, not a virtue. You have to be paid a lot to justify adding it. Most traders are paying the cost without collecting the fee.

Why add complexity if you don’t get paid for it? You have to get paid a lot to add complexity. Otherwise, the complexity is not worth having there.

The mechanism by which complexity destroys you is over-fitting. If your model uses ten indicators that all have to agree, the odds are it works in backtest because you tuned it to the recent past, not because it captures anything real. The more knobs, the more you have curve-fit. The harder it is to execute under stress, the less likely you stick with it when it draws down.

He talks about retail traders contacting him over the website, fixated on the smallest details — exponential versus simple moving average, the exact lookback period, the precise stop multiplier. His answer is always the same: it doesn’t matter. The difference is a rounding error. The fixation on detail is a residue of the technical-analysis culture they grew up in, which sold them on the idea that the right indicator combination is the holy grail.

You won’t really find a hedge fund manager sitting there optimizing his 10 indicators that he’s combining to get to the perfect buying signals. No one really works on that on the institutional side.

The 12-month rule

The simplification he wishes he had put in his first book came from Nicolas Colucci at Quest. The suggestion: forget all the moving-average machinery. Just compare today’s price to the price twelve months ago. Higher means long, lower means short. Two data points. That’s the entire signal.

Clenow ran the math expecting it to be a teaching toy. It wasn’t. Applied to a broad basket of fifty to a hundred futures markets, with standard volatility-based position sizing, it actually outperforms the more elaborate model he published in Following the Trend. He calls this admission “embarrassing.” The downside is that you are in every market at every moment, which eats a lot of margin to equity. But the signal itself does the work.

His advice now is that the 12-month rule should be the benchmark every trend-follower competes against. Build whatever you want, but if you can’t beat that two-data-point rule, you have a problem.

Where the trend-following industry has drifted

Niels (the host, himself a CTA veteran) raises a useful point: return dispersion among trend-following managers has been widening. Clenow’s read is that the universe of pure trend followers has shrunk. Most CTAs are still primarily trend followers, but they have layered on satellite strategies — carry, calendar spreads, counter-trend overlays — to smooth out drawdowns. Some of those overlays helped. Others backfired during specific regimes. That’s where the spread between managers is coming from. The simple trend-following signal still works; what diverges is the bolt-on machinery.

When the rules need a human

Clenow is systematic but not a fundamentalist about it. He brings up the Swiss National Bank de-pegging the franc in January 2015 as an example of where blind obedience to the model gets you killed.

In that case, the franc had been artificially pegged, which meant volatility was suppressed, which meant a standard volatility-based position sizer would tell you to take an enormous position in the EUR/CHF cross-rate future. The model didn’t know it was sitting on a coiled spring. When the peg broke, the people running their systems unsupervised took 25-30% one-day hits.

The lesson he draws is two-fold. First, you need a minimum volatility floor in your position sizer, otherwise artificially low vol regimes will hand you size that bears no relation to actual risk. Second, you need to look at correlated positions. Trading both the Swiss future and the EUR/CHF future was effectively the same bet twice. A trend follower with eyes open would have caught that. The model wouldn’t.

He flags a current example of the same problem: short-term European interest-rate futures trading above 100, where the asymmetry is huge. You can drift up a little, or you can collapse a long way. The model can’t see asymmetry. You have to.

Why stocks are not futures

The second book, Stocks on the Move, was written precisely because trend following doesn’t translate cleanly from futures to equities. The reason is correlation.

In futures, you trade a hundred markets with low internal correlation — bonds, commodities, currencies, indices. If commodities go quiet for two years, currencies might carry you. Diversification is real. In stocks, you have thousands of instruments but they all move together. They look uncorrelated on the upside, but the day the market falls, they all fall together. Your risk model evaporates exactly when you need it.

The second issue: in a bull market every stock gives you a buy signal. You can’t be in all of them. You need a ranking method to decide which subset to own. And the classic trend-following exit — stay in until it stops going up — is wrong for stocks, because in a bull market sideways stocks will sit in your portfolio forever, draining capital while better names rip. So you need a momentum ranking that rotates capital toward the strongest names.

The third issue: a regime filter. Don’t buy stocks in a bear market. If the index is falling, the universe of clean long opportunities collapses, and the model needs to know about it.

Why he doesn’t short stocks

Clenow is blunt: shorting stocks is a fool’s errand for most people. Bear-market stocks behave violently. They drift down for weeks, then have one explosive counter-rally on a takeover bid or a policy intervention, and the short blows up. Shorting anything is harder than going long. Shorting stocks specifically is harder still, especially at the longer horizons where trend following lives.

He does short indices and individual stocks himself, but in size that never makes it the main strategy. The book recommends against it because the book’s audience is retail, and he errs on the responsible side.

Counter-trend, but going the right way

The counter-trend models Clenow likes aren’t the obvious kind — buying after the market has crashed, hoping for the bounce. He doesn’t trust those.

What he likes is counter-trend models that trade in the direction of the dominant trend but enter after a pullback. The genesis was the realization that medium-term trend followers tend to stop out too early — and there’s a fairly predictable distance at which they stop out (three average true range units from the recent extreme, in his first book’s model). So he built an indicator that measures the ATR distance from the recent extreme, and goes the other way exactly where the trend-followers are puking. He published it half-jokingly, given his usual dislike of indicators.

The fund-management bit

This is where the Top Traders Unplugged framing really earns its keep. Niels pulls Clenow into the operational side.

On starting a fund: most people obsess over the revenue side and ignore the cost side. Bloomberg or Reuters is two grand a month. Compliance is now a serious line item in Europe. You will not survive on performance fees — they may or may not come. The base management fee has to cover everything. Which means you need a base of assets large enough that 1-2% covers your salary, your staff, the brokers, the custodian, the administrator. That base is the hard thing to assemble. He got lucky in the mid-2000s, in a regulatory environment that no longer exists.

On regulation: hurdles of entry are climbing fast in Europe. A million-dollar account from a friend is not a business — it’s a future complaint that finds the regulator. He is emphatic about not going under the radar. One bad outcome with an unhappy investor and you’re done in the industry.

On the philosophy of trading other people’s money: his blog post “Why managing your own money is a bad trade” makes the argument that if you’re going to sit at a desk all day trading anyway, you may as well add a fee-paying client book on top of your own capital. The base fee gives you the runway to think long-term. Trading your own money alone forces you into short-term thinking because rent is due at the end of the month. The valid reasons not to take outside money: you genuinely don’t want to deal with people, regulatory drag in your jurisdiction makes the math impossible, or your strategy doesn’t scale.

Why he writes books at all

A question Niels comes back to: why give away the rules? Clenow’s answer is that the “wrong people” — meaning anyone who could actually do damage with them — already know this stuff. The professional quants at every shop have either derived the same ideas or could in an afternoon. And the strategy of becoming a little bit known through the books has paid off in deal flow and client conversations. It’s a business development tool wearing the clothes of altruism.

Key Takeaways

The two-data-point trend-following rule (price today vs. price 12 months ago, long if higher, short if lower) outperforms the more elaborate model in Following the Trend when applied to fifty-plus futures markets with volatility position sizing. Use it as a benchmark — if your model can’t beat it, you have a problem.
Complexity has to earn its keep. Ten indicators that “agree” usually means you’ve over-fit to the recent past. The complexity also makes you less likely to execute the system under stress.
Choice of moving average type, exact lookback, exact stop multiplier — these are rounding errors. The signal class matters, the parameter doesn’t.
A minimum volatility floor in your position sizer is non-negotiable. Artificially low-vol regimes (pegs, central bank ranges) will otherwise hand you ruinous size.
Watch for correlated positions that look like diversification. Trading EUR future and CHF future is the same bet twice.
Asymmetric markets (short-rate futures pinned near zero, pegged crosses) are invisible to trend models. The model can’t see one-way risk. You have to.
Stocks are not futures. Internal correlation is too high — diversification breaks down on the way down. Need a ranking method to pick which stocks to own, and a regime filter to switch off in bear markets.
Shorting stocks: bear-market stocks move violently in both directions. Hard to do well. Most retail traders should not.
The counter-trend model worth having goes with the dominant trend, entering at the point where standard trend-followers are stopping out (around three ATR from the recent extreme).
Trend-follower return dispersion is widening because of differences in the satellite strategies (carry, calendar spreads, counter-trend) bolted onto the core signal — not differences in the core signal itself.
Fund economics: management fees cover the business, performance fees are gravy. Don’t budget the gravy. The hard problem is the initial asset base that lets 1-2% cover salaries and infrastructure.
European regulatory load now makes sub-$10M mandates uneconomic. The window in which Clenow started his first fund no longer exists.

Claude’s Take

The Top Traders Unplugged format pulls something out of Clenow that doesn’t come through in his book interviews or the more famous Quantopian talk: the fund-management arithmetic. Niels is a CTA veteran himself, so the conversation slides easily into base-fee versus performance-fee structure, regulatory drag in Europe, why you take outside money even if you don’t strictly need to. That’s the texture you don’t get from interviewers who treat him purely as a trading-systems guy.

The “simplicity” thesis lands more cleanly here than in his other appearances too. The 12-month rule story is concrete and a little embarrassing for him — he wrote a book showing a model and the most influential lesson turned out to be that an even simpler model works better. That admission gives the simplicity argument credibility it wouldn’t have if he’d led with it.

The weaker stretches are the standard technical-analysis swipes (Elliott Wave, Fibonacci) which are deserved but well-trodden, and a long detour about regulations in Europe that, while accurate, feels like grievance more than insight. The Swiss franc and asymmetric-risk discussion is the standout — it’s the place where a systematic trader makes the case for keeping a human in the loop without it sounding like a cop-out.

Score is 8. The 12-month-rule reveal and the fund-economics section make it more useful than the other Clenow interviews from the same era. Drops a point for the regulatory rant and a slight overlap with material covered better elsewhere on signal mechanics.