Stop-Loss Clustering at Round Numbers: Osler's Mechanism and Index Futures

What Osler's Research Established

Carol Osler's 2003 paper in the Journal of Finance documented a measurable pattern in foreign exchange: stop-loss orders cluster disproportionately at round numbers — figures ending in 00 or 50 and other psychologically salient digits. The clustering is not random. Traders default to tidy numbers when placing orders, and dealers who see aggregated order flow can anticipate where stop-triggered selling or buying will appear. Osler showed that price tends to accelerate after breaching these levels because executing one stop often triggers the next in a cascade, amplifying the move beyond what underlying supply and demand alone would produce.

The mechanism is behavioral at its root. Whether you are a retail trader setting a stop on EUR/USD at 1.1000 or a fund manager protecting an index futures position at a major round figure, round numbers feel precise even when they carry no structural significance. That shared tendency is what creates the clustering — and the cascade that follows a breach.

Why Index Futures Are a Natural Fit for This Framework

Index futures like the E-mini S&P and Nasdaq 100 contract trade around psychologically prominent levels, and participants at every size use those figures as reference points for stops, targets, and position sizing. The same logic Osler applied to foreign exchange round numbers translates: when price approaches a major round figure, resting stop orders from many independent participants pile up just beyond it, and the breach can trigger a self-reinforcing run.

Unlike forex, index futures also carry option-related hedging flow, expiry pinning, and institutional roll activity. These create secondary clustering effects around strikes and roll dates that compound the plain round-number story. The mechanism is plausible — but plausibility is not proof, which is why empirical testing matters.

What 660,005 Backtests Show for the Index Asset Class

IndicatorEdge ran 660,005 out-of-sample backtests across 903 assets — including index instruments — testing 382 indicators on 1-Hour, 4-Hour, Daily, and Weekly timeframes with realistic transaction costs included. Across all assets and timeframes, only 26% of indicator/asset combinations beat a passive buy-and-hold benchmark, and the median Sharpe ratio of the best-performing indicator per asset was 0.62.

For the Index asset class specifically, the indicators that came out on top were DeMarker (2 assets), ROC (30), Ultimate Oscillator, Williams %R (7), and EMA 100 Trend (one asset each). None of those are pure order-flow or stop-hunt tools — they are momentum and oscillator-based signals. The data did not identify a round-number breakout strategy as a consistent winner across index instruments in any of the four tested timeframes.

The closest analog to the Osler stop-cluster concept in the test suite is the SMC: Liquidity Sweep indicator, which targets exactly the stop-cluster-then-continuation setup. Its median win rate across assets was 71.2% — which sounds attractive — but only 8% of the assets it was tested on produced results that beat buy-and-hold. A high win rate paired with poor overall edge is a classic trap: small gains on winners, large losses when the cascade does not materialise.

Why Round-Number Edges Are Hard to Systematize

Osler's findings were robust in the interdealer forex market, where aggregated order book data was available and clustering was measurable. Systematizing the same idea in index futures is harder for several reasons. Stop-loss orders are largely hidden until they execute — you cannot observe the cluster before price reaches it. Algorithmic participants have read the same research and front-run the anticipated cascade, eroding the edge. And not every round number attracts the same density of stops; the ones that matter shift with market context in ways that a fixed rule struggles to capture.

This does not mean the mechanism is inert. It means that turning it into a rule-based indicator that beats buy-and-hold consistently across index assets — across the four tested timeframes — is not what the data shows. The momentum and oscillator signals that did win in the Index class may capture some of the same post-breach momentum, just from a different entry angle.

These Are Hypothetical Backtests — Not Advice

All results on this site, including every figure on this page, are hypothetical backtests. They use out-of-sample data and include realistic transaction cost estimates, but past backtest performance does not guarantee future results. Nothing here constitutes financial advice, a trading recommendation, or a solicitation to buy or sell any instrument. You are responsible for your own trading decisions. Consult a qualified financial professional before acting on any information found here.