Is Trend Still Your Friend? A Microstructural Explanation for the Demise of Short-Term Trend-Following

Trend following is one of the oldest and most persistent anomalies in finance. The evidence that recent winners continue to outperform recent losers has been documented across virtually every liquid asset class, stretching back at least two centuries. It stands in direct opposition to the Efficient Market Hypothesis, yet it has survived out-of-sample testing, multiple market regimes, and decades of institutional capital deployed to exploit it.

However, something changed around 2008-09. A July 2026 paper, “Is Trend Still Your Friend? A Microstructural Account of the Demise of Short-Term Trend-Following,” by Jutta Kurth, Zoltan Eisler, Adam Rej, and Jean-Philippe Bouchaud, takes on a puzzling developments in systematic investing: why short-term trend-following, reliable for decades, essentially stopped working after the Global Financial Crisis, while longer-horizon trend signals kept delivering. Their answer is not what most practitioners have assumed.

The Puzzle

The authors point to the SG CTA Index — the standard industry benchmark for large commodity trading advisors. After roughly a decade of steady gains, the index has been flat to negative for the better part of fifteen years, rescued only by two macro-driven episodes in 2014 and during the Covid shock.

Rebased) SG CTA Index, 2000–2025. Data courtesy of Société Générale Prime Services & Clearing.

The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged and do not reflect management or trading fees, and one cannot invest directly in an index.

The following chart shows the 5-20 day exponentially weighted (more weight given to the more recent days and less to the older days, with the weights decaying exponentially) moving average (EWMA).

The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged and do not reflect management or trading fees, and one cannot invest directly in an index.

Using their own methodological proxy built from about 100 liquid futures contracts across commodities, equity indices, currencies, and government bonds spanning 1995–2025, the authors confirm the same pattern with far more resolution.

Four empirical facts anchor the analysis and constrain any credible explanation:

1. The break is abrupt — a regime shift, not a gradual decay of alpha.
2. It is speed-dependent — fast trend signals (days to weeks) were hit hardest, while slow trend signals were largely unaffected.
3. It is asset-class heterogeneous — equity indices and currencies were hit hard; yields and most commodities were largely spared.
4. There has been no recovery, despite meaningfully improved liquidity and reduced CTA participation since 2018.

Four Candidate Explanations, Three of Which Fail

The authors test four hypotheses for what broke.

Capacity constraints. The idea that too much capital chasing the same signal eroded returns doesn’t hold up on close inspection. CTA industry assets grew through the 2000s and didn’t peak until 2022 — years after trend profit and loss (P&L) had already gone flat, the wrong sequencing for a crowding story. Liquidity in these markets rose sharply after 2018, which should have relieved any capacity constraint, yet performance never recovered. And when the authors recompute returns using same-day execution — effectively removing a full day of market impact and the associated costs — fast-signal performance is still flat after 2008. If capacity and execution costs were the culprit, stripping those costs out should restore profitability. It doesn’t. That’s an important clue: something happened to the underlying signal, not just the cost of harvesting it.

Electronification of futures markets. The shift from floor trading to electronic order books is a popular culprit, but the timing and sequencing don’t match. Electronification was gradual; the P&L break was sudden. Equity futures electronified more than five years before the break; interest-rate futures electronified early, yet never degraded at all.

A shift in CTA-versus-order-flow dynamics. The authors also examine whether a change in how CTA trades interact with aggregate market order flow explains the pattern. It doesn’t map cleanly onto which asset classes degraded and which didn’t.

The Variable That Actually Matters: Volatility-Normalized Tick Size

Having ruled out the usual suspects, the authors identify the variable that cleanly separates contracts whose trend signals survived from those that didn’t: the tick size of a contract, scaled by its volatility. Sorting the futures universe into “small-tick” and “large-tick” halves each month produces a stark dichotomy. Short-term trend P&L on small-tick contracts collapsed to essentially zero after 2008 across every signal speed tested. On large-tick contracts, performance barely budged — continuing to accrue at close to the pre-2009 rate even at the fastest signal horizons.

Notably, sorting by liquidity instead of tick size — despite the two being correlated — does not reproduce this clean separation. Nor does sorting by asset class on its own; the asset-class pattern turns out to be a byproduct of the fact that equity indices and currencies happen to cluster in the small-tick tier, while bonds and most commodities cluster in the large-tick tier.

Two further cuts of the data corroborate the framework. Trend gains have always been concentrated in low-volatility periods (a long-documented pattern known as the LeBaron effect—a negative (inverse) relationship between volatility and serial correlation of returns: when volatility is high, serial correlation tends to be low, and when volatility is low, serial correlation tends to be higher—and small-tick contracts continue to behave like large-tick contracts specifically during those low-volatility stretches — exactly what you’d expect if the ratio of tick size to volatility, not tick size alone, is what matters. And when the data is split by the magnitude of daily returns, the P&L earned on small, quiet trading days has been essentially untouched by the entire episode. It’s specifically the large, high-conviction directional moves — the ones that used to be trend-following’s bread and butter — where profitability vanished on small-tick contracts.

The Mechanism: A Broken Feedback Loop

Here the paper offers its most interesting contribution. The authors argue that trend following isn’t simply a static anomaly that traders harvest — it’s sustained by a self-reinforcing feedback loop. Trend signals prompt directional trades; those trades push prices further in the signal’s direction through market impact; that price movement sustains or strengthens the signal for the next trader to act on. Profitability and the very existence of the trend anomaly are, in this telling, two sides of the same coin. Break the loop, and both profitability and the signal itself should decay together — not just the harvest, but the underlying pattern.

What broke the loop, in the authors’ account, is the post-crisis takeover of market making by high-frequency trading firms operating on tight, flat-inventory mandates rather than the bank-affiliated and proprietary desks of prior decades that were willing to hold overnight risk. HFT market makers are structurally disinclined to absorb the kind of large, predictable, directional flow that CTA rebalancing generates—the literature on this point (HFTs withdrawing liquidity in front of large institutional orders) is well established.

That withdrawal happened across both tick-size tiers, but its consequences diverged sharply based on order book structure. Small-tick order books are inherently sparse — the economics of queue priority mean depth is naturally thin, so when HFT liquidity providers pulled back, there simply wasn’t enough residual depth left for trend followers to execute without materially “walking the book.” Faced with that cost, trend followers appear to have structurally disengaged from fast signals in those contracts — and once that flow disappeared, so did the impact-driven reinforcement that had been sustaining the trend signal in the first place. Large-tick order books, by contrast, are inherently dense, with meaningful resting depth at multiple price levels; residual liquidity there remained sufficient for trend followers to keep executing much as before, the loop kept turning, and both the signal and its profitability survived.

The authors also examined whether switching from aggressive market orders to passive limit orders offers an escape route on small-tick contracts. It doesn’t. A resting limit order tied to a trend signal gets filled preferentially in the wrong states of the world—filled when the market moves against the position, missed when it runs in the signal’s favor—which simply relabels the cost as missed opportunity rather than slippage. And passive orders, by definition, don’t generate the aggressive flow that feeds the self-reinforcing impact loop in the first place. There’s no execution style that sidesteps the underlying microstructural shift.

A Few Caveats Worth Flagging

Before turning to takeaways, as Max Martone noted in his LinkedIn post on the paper, a few caveats are worth naming. Two of the four authors, Rej and Bouchaud, are affiliated with Capital Fund Management, a large systematic manager whose book leans toward diversified, large-tick trend exposure—a microstructural explanation that vindicates that positioning over a capacity/crowding story is, at minimum, convenient for the firm, even if that doesn’t make the finding wrong. The supporting evidence for the mechanism is also less tidy than the headline P&L break suggests: the paper dates the P&L collapse to roughly 2008-09, but the book-imbalance correlation offered as corroboration doesn’t flip until around 2010, the adverse-selection evidence doesn’t flatten until 2011, and the trade-imbalance correlation shift is dated to 2015—four different break points folded into a single causal story. The small-tick/large-tick dichotomy itself is softer than the headline charts imply: the authors’ own robustness checks show the relationship behaves more like a continuous gradient than a clean binary split, meaning some of the crispness in the main results owes to where the median split happens to fall each month. And the central causal claim—that a shift toward HFT market making specifically drove the change in liquidity provision—isn’t directly evidenced in the data; no HFT participants are identified, and the authors acknowledge they can’t distinguish a change in who is providing liquidity from a change in how existing providers behave. None of this overturns the paper’s core empirical finding, which is well documented: small-tick and large-tick contracts diverged sharply after 2008. It does mean the explanation for why is more provisional than the confident framing suggests.

Key Investor Takeaways

The degradation of short-term trend appears structural, not cyclical. The authors see no obvious mechanism by which it reverses absent a genuine change in market-making structure—new regulation, a new class of inventory-tolerant liquidity providers, or a shift in tick-size regimes. That’s a meaningfully different diagnosis than “the trade got crowded and will eventually work again.”

Not all trend is created equal. Longer-horizon trend signals have held up far better than short-horizon ones, and the driver isn’t asset class so much as the underlying market microstructure—specifically, whether a contract’s order book is naturally sparse or dense relative to its volatility. Investors evaluating CTA managers or trend strategies should pay attention to what signal speeds and instrument mix a given strategy relies on, rather than assuming “trend” is a monolithic exposure.

Capacity discussions should happen at a finer level than the industry aggregate. Because the binding constraint operates tier by tier rather than strategy-wide, a fund’s overall participation rate can look comfortably below historical norms while still being effectively capacity-constrained within the small-tick subset of its book.

This is a good reminder of a broader lesson in evidence-based investing: market structure evolves, and strategies whose returns depend on specific liquidity-provision arrangements can see those arrangements change beneath them. The rise of HFT market making has reshaped a lot more than equity market microstructure—this paper is a well-documented example of how it can quietly erode a return premium that had persisted for roughly two centuries.

As always, the paper leaves open questions the authors themselves flag—most notably how much of trend’s historical existence owes to this impact-driven feedback loop versus more traditional behavioral explanations like slow information diffusion or underreaction to news. That distinction matters for how durable slower-horizon trend premiums are likely to be going forward, and it’s an area worth watching as more research emerges.

Larry Swedroe is the author or co-author of 18 books on investing, including his latest Enrich Your Future. He is also a consultant to RIAs as an educator on investment strategies. This article is for informational and educational purposes only and should not be construed as specific investment, accounting, legal, or tax advice.

was originally published at Alpha Architect. Please read the Alpha Architect disclosures at your convenience.