Skip to main content
Track Order flow analysis
6 / 12
Library

2001issue C061-6

Round-number clusters behind support and pattern signals

Historical bank tickets stacked take-profits at whole numbers and 50s and parked stop-losses just beyond those prints. Editorial reading: treat a support line, a head-and-shoulders neckline, and that round-number-cluster as one book, then ask whether the next event is a bounce, a through-level surge, or a crowded pattern that never fills.

  • Take-profits stacked at whole numbers and 50s, while stop-loss buys sat just above those prints and stop-loss sells sat just below, so a crossing could add flow in the same direction.
  • Head-and-shoulders signals on daily exchange rates overlapped simpler moving-average and filter rules, and those simpler rules were correct more often.
  • Daily support and resistance levels from six private firms predicted 15-minute bounce outcomes better than arbitrary numbers, including one dollar-yen record of 26 of 28 monthly windows.
  • Clustered conditional orders create path-dependence: the next increment depends on the exact printed level and can stretch beyond the 15-minute bounce window.
Entries in this reading3 entries

The chart object is a resting book

A support-resistance zone is a published or observed price zone used as a bounce-or-break hypothesis over an intradayan-to-multiweek window. Head-and-shoulders is a three-peak chart structure scored as a position signal and then compared with simpler rules and with actual volume around the entry day. A round-number-cluster is a pile-up of take-profits at whole numbers and 50s, with stop-losses parked just beyond those same prints.

Editorial reading: treat the support line, the neckline, and the round number as one execution problem. First map which resting order type sits at the level. Then decide whether the next event is a bounce, a through-level surge, or a crowded pattern that never fills.

When the neckline already sat in simpler rules

On daily exchange rates spanning about 20 years, head-and-shoulders signals overlapped the positions of simpler moving-average and filter rules, and those simpler rules were correct more often.

On daily prices for 100 randomly chosen equities, the same pattern failed an average-profitability test, yet extra volume peaked on the day a pattern-based position would have been opened and then faded over the next few days.

Editorial: volume that peaks on the entry day can mark a crowded print rather than a leftover ticket that still fills.

Round-number tickets and the same-direction surge

Almost 10,000 stop-loss and take-profit orders from one major foreign-exchange bank, collected over about six months in three currency pairs, showed take-profits stacked at round numbers. Stop-loss buys sat just above whole numbers and 50s, and stop-loss sells sat just below them, so a crossing of the round number could trigger more flow in the same direction.

Resting orders clustered most at prices ending in zero, next at those ending in five, and seldom at endings of one, four, six, or nine.

Order-flow-analysis reads those clustered stop-loss and take-profit tickets, buy-versus-sell initiation, and volume timing to judge what is executable at the level.

Published zones versus arbitrary levels

A bounce was scored if the rate was still above support or still below resistance after 15 minutes, and those frequencies were judged against 10,000 bootstrap sets of arbitrary levels.

Daily support and resistance levels circulated by six private firms over a couple of late-1990s years predicted intradayan turnarounds better than arbitrary numbers. One firm beat that benchmark in 26 of 28 monthly windows on dollar-yen.

Path-dependence after the 15-minute clock

Because the next increment depends on the exact printed level, clustered conditional orders create path-dependence that can stretch beyond a 15-minute window and help explain why short-horizon macro forecasts struggle.

Stop-loss clusters can generate more unusually large moves than expected, while take-profits at round numbers can halt a trend and generate more unusually small moves.

Editorial: a bounce at 15 minutes does not retire stop-loss tickets parked just beyond the same print, so the later event can still be a through-level surge rather than a completed hold.

Educational research material, not investment advice. Historical source context does not establish present-day performance.
6 of 12 in the Order flow analysis track
20011-1 pp.Next on Order flow analysisDisplayed-size filters in multi-venue index-share executionAdvertised bid and offer quantity is a displayed-size-filter for whether an order is executable and what implementation-cost it may carry, not a forecast of the next tick.
All readings on this track · 12 readings
  1. 1988Constructing a value area from TPO, volume, and order flow
  2. 1988Constructing Value Area trends from half-hour auction profiles
  3. 1988Constructing auction structure with volume as a clock
  4. 1999Direct access routing and the visible book
  5. 1999A floor-to-screen case for the trading decision
  6. 2001Round-number clusters behind support and pattern signals
  7. 2001Displayed-size filters in multi-venue index-share execution
  8. 2006A high-volume wide bar as an auction-intent check
  9. 2007Market Profile as auction context for inventory imbalance
  10. 2008Decimalization, liquidity, and why price-only charts miss the tape
  11. 2013Liquidity pockets and order routing after the speed race
  12. 2015Bid-ask queues and the instant-adjustment critique
All 22 readings tagged Order flow analysis
Also on Order flow analysis5 readings