Why round levels hold, and why breaks run — the order book underneath

Osler documented the mechanism from real currency-order data: take-profit orders cluster ON round numbers and stop-losses just BEYOND them. That one asymmetry produces both effects at once — reversals at round levels and faster moves after breaks. This builds a synthetic order book on that asymmetry, runs a random-walk price through it, and scores every approach against matched non-round control levels. Set clustering to zero and the round-number statistic collapses to the control rate, which is the whole point: the effect lives in where orders rest, not in the drawn line.

Resting orders at each level

Take-profit mass ON the level78

Counter-trend limit orders sitting exactly on the round number. They absorb flow and push price back the way it came.

Stop-loss mass BEYOND the level42

Trend-following market orders. Reached, they fire in the direction of the break and push price further on.

Stop-loss offset beyond level12 pips

How far past the round number the stops rest, on both sides.

Cascade feedback0.70

Triggered stop volume adds short-lived momentum, which can reach the next cluster. Emergent, not coordinated — many independent orders firing in waves. 0 = no feedback.

Market & clustering

Price volatility4.0 pips/tick

Clustering intensity80

0 = the honest control case. Identical total order mass, spread uniformly across every price, so no round number holds more than anywhere else. The round-number statistic must then vanish.

Level grid

Sample set#1

Step through one approach

Ticks revealed1500

Step forward and watch one approach in slow motion: the ladder below drains as price consumes the resting orders, then refills as new ones arrive.

Step size10 ticks

Bounce rate at round levels minus matched non-round control levels

percentage points, synthetic data — not a measured market statistic

Round-level approaches

Bounce rate, round levels

Bounce rate, control levels

Move in 20 ticks after a break vs a bounce

mean absolute, pips

Stop volume triggered

order units, all samples

Difference in standard errors

Independent 1,500-tick synthetic samples pooled for these statistics40

Sample 1 · synthetic price against the round-number grid

Synthetic price Round level Non-round control level Reversed off a take-profit cluster Broke through, triggered stops

take-profits ON the level → price pushed back · stop-losses N pips BEYOND → price pushed onward

Order-depth ladder · remaining mass at the tick shown

Take-profit mass, on the level Stop-loss mass, beyond the level Full replenished size Current price

Bounce rate: round levels vs matched non-round control levels

Two things this is built to show you. First, drag clustering intensity to 0: the order mass is identical, only spread uniformly across every price instead of piled at round numbers, and the round-level bounce rate collapses onto the control rate. The effect lives entirely in where orders actually rest, not in the line drawn on the chart. Second, push stop-loss mass up: the bounce rate at round levels falls below the control rate while post-break moves get much larger. Because take-profit clusters reverse price and stop clusters extend it, a “sweep” is not a one-way trade — which is exactly why the retail stop-hunt story does not follow from the order-flow evidence. This is a mechanism demonstration on synthetic data: it reproduces no paper’s measured magnitudes, and there is no strategy, position, or account here.

What the research actually found

Support level: partial. The order-flow mechanism is documented. The trading narrative built on top of it is not.

“Support/resistance levels published by major FX dealers had genuine intraday predictive content — prices bounced off them more often than chance (Osler 2000). The mechanism is documented: take-profit orders cluster ON round numbers and stop-losses just BEYOND them, so trend reversals at round levels and faster moves after breaks are real order-flow effects, not chart mysticism (Osler 2003).”

“Round-number price clustering is one of the most replicated microstructure findings. Stop and take-profit orders cluster at and just beyond those levels, and reaching a stop cluster can trigger positive-feedback ‘price cascades’ that make moves unusually fast. So ‘liquidity rests at obvious levels, and a move into it can fuel a fast continuation’ is genuinely real and measured.”

What is not supported.

“The idea that institutions deliberately drive price to a level to trigger retail stops and then reverse. The measured cascades are emergent — many independent orders firing in waves — not a coordinated raid; predatory-trading theory concerns forced liquidation of large distressed positions, not retail stops. And take-profit clusters reverse price, so a ‘sweep’ is not a one-way trade. Most direct evidence is FX from ~1996–2005.”

“Predictive content is not a trading system: the documented effects are intraday, small, and strongest in FX where the order data lives. Nothing supports precise ‘retest’ entry rituals or the idea that drawn lines on any chart carry power beyond where orders actually cluster.”

So: round-number clustering and the cascades that follow from it are real, measured, intraday and small, and the direct evidence is largely FX from about 1996–2005. The deliberate stop-hunting raid is the part with no support behind it.

Harris, Lawrence (1991). “Stock Price Clustering and Discreteness.” Review of Financial Studies, 4(3), 389–415. DOI: 10.1093/rfs/4.3.389.

Osler, Carol L. (2000). “Support for Resistance: Technical Analysis and Intraday Exchange Rates.” FRBNY Economic Policy Review, 6(2), 53–68.

Osler, Carol L. (2003). “Currency Orders and Exchange Rate Dynamics: An Explanation for the Predictive Success of Technical Analysis.” Journal of Finance, 58(5), 1791–1820. DOI: 10.1111/1540-6261.00588.

Brunnermeier, M. K., & Pedersen, L. H. (2005). “Predatory Trading.” Journal of Finance, 60(4), 1825–1863. DOI: 10.1111/j.1540-6261.2005.00781.x.

Osler, Carol L. (2005). “Stop-loss orders and price cascades in currency markets.” Journal of International Money and Finance, 24(2), 219–241.

Bhattacharya, U., Holden, C. W., & Jacobsen, S. (2012). “Penny Wise, Dollar Foolish: Buy-Sell Imbalances On and Around Round Numbers.” Management Science, 58(2), 413–431. DOI: 10.1287/mnsc.1110.1364.