Liquidity Sweeps and Stop Runs in Futures Trading: How to Spot and Trade Them

You short near the prior day high, your stop sits just above it, price pokes through for a few ticks, you're out, and then ES drops the way you expected. That sequence is common enough that traders have names for it: sweep, stop run, stop hunt. The pattern is real on charts. Why it happens, and how reliably it can be traded, is much less certain than most write-ups claim.

The mechanism, and what we can't know

Plenty of traders place stops just beyond obvious levels like the prior day high or low, a round number or the weekly extreme. A stop to buy becomes a market order when triggered. When price trades through a crowded level, those orders fire and add buying that wasn't there a moment before. A large seller can find that liquidity useful. That's the standard explanation.

What you can't tell from a chart is whether anyone intended it. A level can be run because resting orders sat behind it, because a news item hit, or because a big order needed to be filled. The pattern is observable. The motive isn't, and I'd avoid building a trade thesis on a story about who is hunting whom.

What a sweep looks like

1. Price consolidates just under a widely watched level.
2. A fast bar pushes a few ticks through it, often with elevated volume and a long wick.
3. One to three bars later price is back on the original side of the level.
4. Follow-through away from the level either appears or doesn't. If it doesn't, the trade was just a failed breakout fade.

How long this takes varies, so don't set a fixed clock on it.

Sweep or real breakout

Leaning toward a sweep: a single spike with a long wick, a close back inside the range, volume that spikes and fades, no catalyst, and an obvious level that many traders watch. Leaning toward a real breakout: multiple closes beyond the level, volume that holds up, a scheduled catalyst, or a level that has been tested repeatedly.

The cleanest filter is a close. If a 5-minute bar closes well beyond the level, treat it as a breakout until price proves otherwise. The cost is that you'll enter later and give up some of the move, and sometimes price closes beyond the level and still reverses. Choose the timeframe in advance. Avoid fading anything around a major release.

Opens and closes tend to be active periods, and spreads and slippage can be worse then. Label all times in Eastern Time if you plan around them, and check holiday and early-close schedules.

A priced example

This is hypothetical. The prior day high is 5,250.00 and ES has been compressing between 5,235.00 and 5,250.00. A spike takes the high to 5,252.00 (8 ticks past the level) and the bar closes back at 5,249.00.

That's 3.5 times the risk before fees. On an MES contract the numbers are one tenth: $20 risk, $70 reward. Real fills will differ, and a stop in a fast market can slip. If the stop is farther than your risk plan allows, reduce size or skip. Don't enter the spike bar itself. Waiting for the reversal bar to close costs you part of the move but removes some failures where price just keeps going.

Levels to mark before the session

• Prior day high and low. Decide whether you mean regular session or full session, and stay consistent.
• Prior week high and low.
• Round numbers inside the likely range.
• The opening range high and low once it forms, after the first 15 to 30 minutes of RTH.

No level is guaranteed to produce a sweep, and I'm not quoting frequency. A level gets run, or it doesn't.

Where YMI's KPL fit

YMI's KPL are fixed reference prices anchored to the RTH regular-session opening print, with four levels in the standard ES and NQ map that don't move intraday. They aren't known before the open. If a sweep happens near one, mark it on your chart and journal the outcome. I won't claim any bot is designed to trade sweeps. For what the automated variants actually do, read the KPL bot specs and the Marty bot specs. Level construction is on the KPL trading page.

Testing the idea

Write a rule for what counts as a sweep: level, minimum and maximum penetration in ticks, close requirement, entry and stop. Mark up past sessions before looking at what happened next, then record every occurrence, including the ones that failed. Keep a column for whether a scheduled release was involved. Thirty examples help you refine the definition but don't establish an edge, and simulated fills overstate how well you'd have done at the extreme.