Why do stops cluster — and get hit?

Traders complain that the market "hunted" their stop: price dipped just below an obvious level, triggered the exit, then reversed and ran without them. You don't need to believe anyone was hunting you to explain it. The geometry alone is enough.

Everybody reads the same chart

Where do stops go? Just beyond the obvious: below the visible swing low, under the round number, past the moving average, outside the pattern boundary. The obvious levels are obvious to everyone, which means the stops pile up in the same few ticks.

Now recall what a resting stop-sell is: a market sell waiting to happen. A cluster of them is a pool of guaranteed selling that activates at a known price. And the activation is self-reinforcing — the first tranche of triggered stops pushes price into the next tranche.

Two documented market facts make this concrete rather than folkloric: prices and orders cluster at round numbers across essentially every market that has been studied, and participants who need to trade in size prefer to transact where liquidity is thickest. Predictable liquidity attracts flow. That is not a conspiracy; it's the definition of a market.

The trade-off, priced out

Support sits at the round $50.00, with the visible swing low at $49.85. ATR(14) is $1.50. A hypothetical entry at $51.00 with a $500 risk budget:

The buffer costs 20% of the position size and buys immunity to a thirty-cent poke through the level. Same $500 at risk either way. That's the actual decision — not "how do I avoid getting hunted?" but "what am I willing to give up in size to stop being where everyone else is?"

A thirty-cent poke looks like this. Nobody had to be hunting anybody:

The three honest options

1. Buffer beyond the obvious level and accept a smaller position.
2. Keep the tight stop and accept a higher rate of being stopped out on noise — which is survivable if the payoff on the trades that work is large enough (Unit 3 does that arithmetic).
3. Wait for the sweep, entering only after the level has been probed and rejected — which trades a better entry for a worse hit rate on setups that never get probed.

None is free, none is right, and anyone selling one as the answer is skipping the trade-off. The point of this lesson is that the trade-off is computable, and the table above is what computing it looks like.

In the data

Why the numbers coincide is arithmetic, not telepathy. The classic pivot set is a fixed function of one bar: (high + low + close) ÷ 3 for the pivot, 2 × pivot − low for the first resistance and 2 × pivot − high for the first support. Here is the bar Apple printed on 2 January 2024:

Everyone who runs it gets 185.99, 188.09 and 183.54, to the cent, and anyone using those pivots to place a stop "just under support" for the next session was placing it just under 183.54.

Try it now

1. Read the marked swing low and the depth of the sweep beneath it off the schematic above. Then say which of the two stops in the table survives that bar, and what the survivor cost in shares.
2. Now on real prices. In the year below, find a stretch trading within a few percent of a round number, locate the most recent swing low beneath it, and compute the share count for a hypothetical $35,000 account at 1% risk — first with the stop just under that low, then with a buffer of 0.25 × Apple's latest 14-day ATR, which is in the table under the chart.

3. Write the trade-off as one sentence with two numbers in it: how many shares you gave up, and how many cents of noise you bought immunity to.