For about fifteen years now, one of the most reliable strategies in systematic investing has quietly fallen apart - but only at one end of the time spectrum. Short-term trend following, the engine that powered the CTA industry for decades, has been generating returns close to zero since around 2009. Trend signals that operate over weeks to months still work. A new paper by Jutta G. Kurth, Zoltan Eisler, Adam Rej, and Jean-Philippe Bouchaud finally gives us a clear, precise explanation for why the short end broke - and it's not what most people in the industry expected.
The Break No One Fixed
The SG CTA Index climbed steadily through the 1990s and 2000s, then stalled around 2008/9. There were two brief exceptions - one around 2014 and another during Covid - but otherwise the strategy has barely moved. What makes this so striking is that the recovery most people predicted never came.
CTA assets under management topped out around 2012, well after returns had already flattened. Futures market liquidity improved sharply from 2018 onward, which should have helped by reducing the industry's market footprint. It didn't. The break was sudden. It was tied to signal speed - fast signals (days to weeks) broke hardest, while slower ones mostly survived. And there's no sign it's healing.
The Loop That Made Trend Work
To understand what broke, you first need to understand what made trend following work in the first place. The authors describe a self-fulfilling feedback loop:
"Trend signals trigger directional trades, whose market impact reinforces the very price moves that produced the signal, which in turn sustains the signal for the next trader to act on."
— Kurth, Eisler & Rej
In plain terms: trend following didn't just exploit price momentum. It helped create it. Trades generated by trend signals moved prices, which strengthened the signals, which triggered more trades. The strategy's profits and its very existence were driven by the same feedback mechanism. Break the loop, and you don't just hurt returns - you eliminate the signal itself.
Three Explanations That Don't Hold
The paper works through four candidate explanations. Three don't survive scrutiny.
The most intuitive one - that CTAs simply got too big and crowded themselves out - doesn't add up. CTA participation in futures markets is estimated below 1% of total exchanged volume. Running the math on market impact at that scale produces a Sharpe drag of maybe 0.1 per year. The observed collapse was far larger. And here's the telling detail: even when researchers recalculated returns using the same day's closing prices - removing a full day of execution impact from the equation - post-2008 returns were still flat. The signal had gone, not just the economics.
Electronification - the shift to electronic trading - also fails to fit the facts. Some markets went electronic before the break with no ill effects. Others electronified early but showed no degradation. The shift to electronic trading was gradual; the break was abrupt. And order-flow data tells contradictory stories: commodities show a clear regime change in order flow patterns, yet trend profits in commodities held up just fine.
The Tick That Tells the Story
What actually explains it? The authors identify one variable that cleanly separates the strategies that survived from those that didn't: the volatility-normalised tick size. Think of it as how large the minimum price movement is relative to how much a market typically moves in a day.
Split roughly 100 futures contracts each month into small-tick and large-tick groups and the pattern is stark. Post-2008 trend profits have essentially vanished for small-tick contracts, across every signal speed. Large-tick contracts, as Kurth, Eisler, and Rej put it, are "virtually unaffected by the break and continue to accrue at roughly the pre-2009 rate, even at the highest frequency."
Pre-break Sharpe ratios clustered around 0.8 for small-tick contracts and 1.4 for large-tick. Post-break, small-tick Sharpes dropped to near zero. Large-tick held in the 1.0 to 1.2 range. The asset-class patterns most observers noticed - equities and currencies suffering, bonds and commodities not - turn out to be a side effect. Equity indices and currencies tend to be small-tick markets. Government bonds and most commodities tend to be large-tick.
HFT and the End of the Loop
Why does tick size matter so much? The answer runs through the transformation of futures market making after 2008. Traditional market makers - mostly bank-affiliated desks - were replaced by high-frequency trading firms. And HFT market making, the authors argue, is "structurally incompatible with absorbing the predictable and persistent directional flow that aggregate CTA trading generates."
The key difference lies between thin and thick order books. Small-tick markets have thin books: to get to the front of the queue, you have to offer a better price, posted volume is low, and gaps between price levels are common. Large-tick markets have thick books: the spread is wide enough to compensate for risk, and volume stacks up across multiple levels.
When HFT firms detect that a wave of trend-following orders is coming, they pull their bids or offers. In a small-tick market, that leaves almost nothing behind. In a large-tick market, enough volume remains that trades go through at workable prices and the loop continues. The feedback mechanism survives in one environment and collapses in the other.
The data makes this visible: adverse selection for liquidity providers in small-tick futures has "nearly entirely disappeared" since 2011. Depth simply stopped resting long enough to be run over.
Nowhere to Hide
The obvious workaround would be for trend followers to switch from aggressive market orders to passive limit orders, sidestepping the HFT withdrawal problem entirely. The authors close this door too.
A trend follower using limit orders stops adding to the directional flow that moves prices. And directional flow moving prices is the whole mechanism. As the paper puts it: "a trend follower who replaces market orders with limit orders ceases to contribute to the trade imbalance in the direction of the trend, and no longer participates in the impact-mediated reinforcement of the very price moves they are trying to exploit. The loop, on which trend depends both for its profitability and for its existence, is broken from the trend follower's side."
This is a structural dead end. Not a temporary problem, not a market phase - a permanent feature of how these markets now work.
5 Key Takeaways for Advisors & Investors
1 Short-term trend following is structurally broken - not just unlucky. Fifteen years without recovery, even as liquidity improved and industry participation fell, rules out crowding or bad timing as the explanation. The underlying mechanism has changed.
2 Tick size - not asset class - is what separates the winners from the losers. The volatility-normalised tick size cleanly predicts which markets still generate trend profits and which don't. Large-tick contracts (government bonds, many commodities) continue to perform. Small-tick contracts (equity indices, currencies) have effectively stopped.
3 The signal has decayed - not just the trade execution. This isn't just about higher transaction costs eating into returns. The trend signal in small-tick markets has itself deteriorated. The feedback loop that generated and sustained the signal is gone.
4 HFT market making is the structural cause. The post-crisis shift to high-frequency trading firms as the dominant providers of market liquidity is the mechanism behind the break. These firms are not designed to absorb predictable, persistent directional flow. That's the equilibrium state of the market now.
5 Slower signals and large-tick exposure are where alpha still lives. Trend signals at horizons of weeks to months remain meaningful. CTAs that have preserved performance have likely done so through greater exposure to large-tick contracts and longer signal horizons. When evaluating managers, understanding which end of the time spectrum they operate on is due diligence, not a detail.
Footnote:
1 Kurth, Jutta G., Zoltan Eisler, Adam Rej, and Jean-Philippe Bouchaud. "Is Trend Still Your Friend? A Microstructural Account of the Demise of Short-Term Trend-Following." arXiv, 2 July 2026, https://arxiv.org/html/2607.01550.
