💰 quantitative finance

Is Trend Still Your Friend?: A Microstructural Account of the Demise of Short-Term Trend-Following

This paper argues that the post-2009 collapse of short-term trend-following profitability is driven by a microstructural shift where high-frequency market makers' liquidity withdrawal in small-tick contracts disrupts the self-fulfilling feedback loop between trend signals and price impact, whereas large-tick contracts remain insulated due to sufficient residual order book depth.

Original authors: Jutta G. Kurth, Zoltan Eisler, Adam Rej, Jean-Philippe Bouchaud

Published 2026-07-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Jutta G. Kurth, Zoltan Eisler, Adam Rej, Jean-Philippe Bouchaud

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: The "Magic" That Stopped Working

For over 200 years, a popular investment strategy called Trend Following worked like magic. The idea was simple: if a price has been going up, buy it; if it's been going down, sell it. Historically, this worked because the act of buying pushed the price up even further, and selling pushed it down, creating a self-fulfilling cycle that made money for everyone involved.

However, around 2009, something strange happened. This strategy stopped making money, but only for certain types of investments and only for fast trading (looking at trends over days or weeks). Slow trading (looking at trends over months) still works fine.

The authors of this paper investigated why this happened. They ruled out common excuses like "too many people are doing it" (capacity) or "computers took over" (electronification). Instead, they found the culprit lies in the size of the price steps (called "tick size") and how modern "market makers" (the people who provide liquidity) behave.

The Core Concept: The "Self-Fulfilling Loop"

To understand the problem, you have to understand how the strategy used to work. The authors describe it as a Self-Fulfilling Loop:

The Signal: A computer sees a price rising.
The Trade: The computer aggressively buys the asset.
The Impact: Because the computer buys a lot, the price is forced up even more.
The Reinforcement: That new, higher price triggers more computers to buy.

The Analogy: Imagine a crowd of people pushing a heavy car.

Before 2009: The car was on a smooth, wide road. When the first group pushed, the car moved easily. That movement encouraged the second group to push harder, and the car kept rolling. The "push" (the trade) created the "movement" (the trend).
After 2009: On some roads, the surface changed. When the first group pushed, the car didn't move. The second group saw it wasn't moving and stopped pushing. The loop broke.

The Real Culprit: "Small-Tick" vs. "Large-Tick" Roads

The paper discovered that the strategy didn't fail everywhere. It failed specifically on contracts where the price moves in tiny steps (Small-Tick). It still works on contracts where the price moves in large steps (Large-Tick).

Small-Tick Contracts (The Broken Loop): Think of these as a narrow, crowded sidewalk. The price steps are tiny (like pennies).
Large-Tick Contracts (The Working Loop): Think of these as a wide, open highway. The price steps are larger (like dollars).

Why did it break on the "Sidewalks"?

The paper argues that after the 2008 financial crisis, the "market makers" (the people providing the liquidity to let you buy and sell) changed their behavior. They switched from being "inventory-tolerant" (willing to hold stock and absorb big swings) to High-Frequency Traders (HFTs) who are very risk-averse.

The "Ghosting" Effect:

On the Highway (Large-Tick): There is so much depth and space that even if HFTs pull back a little, there is still enough room for the trend-followers to push the car. The loop keeps working.
On the Sidewalk (Small-Tick): The road is already narrow. When HFTs see a predictable wave of buyers coming (the trend followers), they get scared and pull their liquidity away instantly. They stop offering to sell.
- The trend followers try to buy, but there is no one to sell to.
- To get the trade done, they have to "walk the book" (jump up many price steps), which costs too much money.
- Because they can't trade cheaply, they stop trading.
- Because they stop trading, the price doesn't move up.
- Because the price doesn't move up, the "signal" that told them to buy disappears.

The Loop is Broken: The trend followers didn't just stop making money; the signal itself stopped working because the mechanism that used to create the price movement (their own aggressive buying) was severed.

Why Other Explanations Failed

The authors tested other theories, but they didn't fit the facts:

"Too many people" (Capacity): No, because the money stopped working before the amount of money in these funds peaked.
"Computers took over" (Electronification): No, because some markets went electronic years before the strategy broke, while others broke later.
"Liquidity dried up": No, because total liquidity actually increased after 2010, but the strategy still failed on small-tick contracts.

The "Passive" Trap

You might ask: "Can't trend followers just stop buying aggressively and wait for the price to come to them (using limit orders)?"

The paper says no, for two reasons:

The "Missed Train" Cost: If you are waiting to buy a rising stock, and you place a passive order, the price will likely zoom past your order before you get filled. You miss the profit.
Breaking the Loop: Even if you got filled, passive buying doesn't push the price up. Remember, the trend only works because the buying pushes the price up. If you stop pushing, the car stops moving, and the trend dies.

Summary

The "Trend Following" strategy didn't die because of too much competition or bad data. It died on small-tick contracts because the modern "market makers" (HFTs) became too skittish to support the heavy, predictable buying pressure of trend followers.

On wide roads (Large-Tick): The traffic flows, the car moves, and the strategy works.
On narrow sidewalks (Small-Tick): The traffic police (HFTs) shut down the road the moment they see the convoy coming. The convoy stops, the signal vanishes, and the strategy collapses.

The paper concludes that this is a structural change in how markets work, not a temporary glitch, and it specifically targets the "fast" trends on assets with tiny price steps.

Technical Summary: "Is Trend Still Your Friend? A Microstructural Account of the Demise of Short-Term Trend-Following"

Problem Statement
Systematic trend-following strategies have historically been a robust source of alpha, with empirical evidence spanning nearly two centuries. However, since approximately 2008/2009, short-term trend signals have ceased to deliver reliable returns, particularly within the Commodity Trading Advisor (CTA) industry. While the aggregate performance of the SG CTA Index has flattened or turned negative post-2008, the decay is not uniform. It exhibits specific stylized facts:

Abruptness: The break resembles a structural regime shift rather than gradual alpha decay.
Speed Dependence: Fast signals (horizons of days to weeks) are severely degraded, while slow signals remain relatively intact.
Asset-Class Heterogeneity: Equity indices (IDX) and currencies (FXR) have seen trend profitability vanish, whereas yields (YLD) and most commodities (CMD) have not.

The paper seeks to identify the root cause of this divergence, rejecting standard explanations such as capacity constraints, market electronification, or simple order-flow shifts, in favor of a microstructural mechanism rooted in the volatility-normalised tick size.

Methodology
The authors analyze a cross-section of approximately 100 liquid futures contracts (1995–2025) across four sectors: commodities, equity indices, currencies, and government bonds.

Signal Construction: They employ Exponentially Weighted Moving Average (EWMA) signals (EWM- $\tau$ - $4\tau$ ) normalized by volatility.
Portfolio Construction: Two portfolios are constructed: an "equal-risk" portfolio (volatility-normalized) and a "liquidity-weighted" portfolio scaled to match the aggregate risk of the CTA industry (using a proxy derived from BarclayHedge AUM data).
Diagnostic Variables: The study evaluates four candidate hypotheses (H1–H4) by testing their alignment with the timing, magnitude, and cross-sectional patterns of the PnL decay.
- H1: Capacity constraints (crowding).
- H2: Electronification of markets.
- H3: Structural shifts in CTA vs. order-flow interactions.
- H4: A microstructural mechanism involving liquidity withdrawal.
Key Metric: The authors introduce the volatility-normalised tick size ( $\bar{\rho} = \Psi / \sigma$ ) as the primary cross-sectional variable. Contracts are partitioned monthly into "Small Tick" (ST) and "Large Tick" (LT) tiers based on this ratio.

Key Findings and Results

Rejection of Alternative Hypotheses:
- Capacity (H1): CTA Assets Under Management (AUM) continued to grow after the PnL break, and liquidity increased post-2018 without a corresponding recovery in trend returns. Furthermore, removing execution costs (zero-lag execution) does not restore profitability, indicating the signal itself has degraded, not just the cost of harvesting it.
- Electronification (H2): The transition to electronic trading was gradual and varied by sector, failing to match the abrupt, asset-class-specific nature of the PnL break.
- Order Flow (H3): While correlations between CTA trades and book imbalances shifted around 2010, this shift does not map cleanly to asset-class performance (e.g., commodities show correlation shifts but no PnL decay).
The Central Empirical Result (Tick Size Dichotomy):
- When the universe is partitioned by volatility-normalised tick size, a stark dichotomy emerges.
- Small-Tick Contracts: Post-2008, short-term trend PnL has collapsed to near zero across all signal horizons. The Sharpe ratios for fast signals dropped from ~0.8 to ~0.1.
- Large-Tick Contracts: Trend profitability remains essentially intact, with Sharpe ratios for fast signals staying in the 1.0–1.2 range.
- This partition explains the data better than asset class or liquidity. The apparent asset-class heterogeneity (e.g., FX/Equities failing while Yields/Commodities survive) is a downstream consequence of these asset classes clustering in the Small-Tick and Large-Tick tiers, respectively.
Mechanism: The Broken Self-Fulfilling Loop:
The paper posits that trend profitability relies on a self-fulfilling feedback loop:
- Signal $\to$ Trade $\to$ Market Impact $\to$ Reinforced Signal.
- Trend followers execute aggressively; their market impact pushes prices in the direction of the signal, reinforcing the signal for subsequent traders.
- The Break: Post-2008, the market structure shifted toward High-Frequency Trading (HFT) dominated market making. HFTs, operating on flat-inventory mandates, systematically withdraw liquidity in the face of predictable, large institutional directional flow (CTA trades).
- Asymmetry:
  - Small-Tick (Sparse LOB): The limit order book is intrinsically sparse. HFT liquidity withdrawal removes the residual depth required for aggressive execution. Trend followers are forced to "walk the book," incurring prohibitive costs, or disengage entirely. Once they disengage, the impact-mediated reinforcement vanishes, and the signal itself decays.
  - Large-Tick (Dense LOB): These books possess sufficient residual depth at multiple levels. HFT withdrawal does not prevent aggressive execution. The loop remains intact, preserving both the signal and the PnL.
Price Impact Evidence:
- Analysis of 5-minute returns vs. book imbalance shows that pre-2011, small-tick markets exhibited adverse selection (liquidity providers lost money to trend flow). Post-2011, this adverse selection correlation flattened to near zero.
- This flattening indicates that liquidity providers (HFTs) no longer rest depth in front of predictable flow; the depth is withdrawn before it can be "run over." The cost has migrated from the market maker's inventory loss to the trend follower's slippage and missed opportunities.
- Passive execution (limit orders) is not a solution: it incurs "missed opportunity" costs (failing to fill as price runs away) and, crucially, forfeits the self-reinforcing impact channel because passive orders do not generate the aggressive flow necessary to sustain the trend.

Significance and Claims
The paper claims to provide a microstructural explanation for the demise of short-term trend following, moving beyond capacity or behavioral arguments. Its primary contributions are:

Identification of the Discriminant: Establishing that volatility-normalised tick size, rather than asset class or liquidity, is the critical variable determining trend survival.
Mechanism of Decay: Demonstrating that the decay is structural, caused by the severing of the self-fulfilling impact loop due to HFT liquidity withdrawal in sparse (small-tick) order books.
Signal vs. Harvest: Arguing that the signal itself has degraded, not just the strategy's execution costs. The "harvest" failed because the "crop" (the price move) no longer grows due to the broken feedback loop.
Implications: The break is structural and unlikely to reverse without a change in liquidity provision (e.g., new inventory-tolerant intermediaries). It suggests that capacity constraints should be estimated at the tick-size tier level, as aggregate participation rates can mask binding constraints within the small-tick subset.

The authors maintain modesty regarding the exclusivity of their mechanism, acknowledging that other factors (information diffusion, behavioral underreaction) likely coexist and contribute to trends, but arguing that the impact-loop component has been disproportionately weakened on small-tick contracts, driving the observed collapse.