What really sits at a price level: orders, clustering and stops

The previous lesson set out the supply and demand method on its own terms: a large order cannot fill at one price, the unfilled remainder is left at the base, and price reacts when it returns. That story makes three separate claims, and they do not stand or fall together.

1. Large orders cannot be filled at a single price.
2. The unfilled part stays parked at the original price.
3. Returning to that price produces a reaction.

This lesson tests each against research on how orders actually reach a market. The order book itself, the list of resting buy and sell limit orders that every price is formed from, is covered in the market microstructure lesson on the limit order book, and is assumed here.

Key idea: A method can be right about what happens and wrong about why. That combination is common in trading, and it matters, because the wrong explanation predicts different things in the cases that decide whether you make money.

The first claim is correct, and well documented. An institution that needs to buy far more than the book holds at the best price does not send one order. It sends a metaorder, a parent instruction executed as a long sequence of small child orders spread over minutes, hours or days.

The reason is cost. Buying everything at once would walk up through the book and pay progressively worse prices. Market impact research finds that the price move caused by a metaorder grows roughly with the square root of its size, so splitting and pacing the order reduces the total paid.

So the method's premise is sound: large participants cannot fill at one price, and the footprint of their buying is real.

Definition: A metaorder is the full quantity a trader intends to buy or sell. Child orders are the individual pieces sent to the market to execute it.

The second claim is where the story departs from how execution works. Institutional orders are mostly worked by algorithms designed to hide size and to keep executing as the market moves.

Only the last case matches the method's picture, and it is one execution style among several. A schedule-driven algorithm that has not finished simply keeps buying at higher prices. Nothing on a price chart tells you which kind of order produced a base.

Gotcha: The method reads the chart as if every base were built by a price-capped buyer who stopped when price left. That is possible, but it is an assumption, and it is not the typical case.

The third claim, that price reacts at certain levels, is supported by research. The mechanism that research found is not leftover institutional orders. It is where ordinary traders put their exit orders.

Carol Osler, then at the Federal Reserve Bank of New York, examined the stop-loss and take-profit orders held by a large foreign exchange dealing bank. Her 2003 paper in the Journal of Finance found that the prices on these orders were strongly clustered at round numbers, and that the two kinds clustered differently:

• Take-profit orders clustered at round numbers. A trader who is long and wants to bank a gain picks 1.2000, not 1.1987.
• Stop-loss orders clustered just beyond round numbers, so that a stop sits on the far side of the level the trader expects to hold.

That asymmetry is the whole mechanism, and the next step shows what it does to price.

Price rising toward a round number, with long traders' take-profit sells resting at it and short traders' stop-loss buys resting just above it.

Illustrative numbers, showing why the break is so much faster than the stall. The sell side of a book just above 1.2000, with stop-loss buy orders waiting to become market orders:

Buyers must absorb all 40 at 1.2000, which takes time: the stall. Once that is gone, one more buyer lifts the 5 at 1.2002. That triggers 30 stop buys, which need 30 lots of supply but find only 5 at 1.2004, so they sweep it. Reaching 1.2004 triggers 25 more. Price covers three levels almost instantly, carried by orders nobody chose to place at that moment.

Osler described this pattern as a price cascade in a 2005 paper in the Journal of International Money and Finance, and it is why broken levels tend to run.

Round-number clustering is not peculiar to currency dealers. Bhattacharya, Holden and Jacobsen, writing in Management Science in 2012, examined a random sample of more than 100 million stock transactions.

They found excess buying by traders demanding liquidity at prices one cent below round numbers, and excess selling one cent above them. The imbalance was largest next to whole dollars, smaller next to half dollars, and smaller again next to quarters: the rounder the number, the bigger the effect.

In practice: This is a direct, large-sample measurement of what the zone method assumes without measuring. Traders really do concentrate orders at particular prices. The prices they choose are just set by how round a number looks, not by where an institution happened to leave a base.

There is a second, simpler reason some levels hold: enough traders expect them to.

If many participants mark the same price, as a zone, a support line or a round number, they place limit orders and stops there. Those orders are real, and they produce exactly the stall and cascade described above. The level matters because it is watched, and it is watched because it has mattered.

Zone traders are part of this. The previous lesson noted that the method's entry is a resting limit order at the zone edge. Every trader following the method adds orders to the zone they drew.

Set side by side, the method's story and the research agree more than a sceptic expects and less than a practitioner claims.

The practical difference is where the explanation leads. The unfilled-orders story points you to the base of a sharp move. The clustering evidence points you to round numbers and to levels many others are watching, which may or may not coincide with a base.