When EURUSD Sweeps a High, Does It Reverse? Usually Not.
Reversal rate (%)
Trading social media loves the "liquidity sweep": price spikes beyond a swing high, grabs the stops resting above it, then reverses. It's a compelling story. We tested it mechanically across every EURUSD occurrence in 16 years — a wick beyond a recent swing high or low that closes back inside — and measured what happened over the following day. 4,986 sweeps.
How we tested it
This public study is derived from OpenAlpha's internal TradingDesk EURUSD Sweep Reversal Atlas. A qualifying event is an H1 candle that wicks beyond an H4 swing high or low identified from the preceding 50 H4 candles, then closes back inside that level. We measure the outcome over the following 24 H1 candles:
- Continuation: an H1 candle closes beyond the swept level.
- Reversal: price moves one H4 ATR away from the level without an H1 close beyond it.
- No clean resolution: neither condition occurs in the measurement window.
This is a historical classification study, not a forecast or a standalone trading signal.
The story is mostly wrong
| Outcome | Count | Share |
|---|---|---|
| Continuation (price closes beyond the level) | 3,402 | 68.2% |
| Reversal (rejects and moves away from the level) | 1,506 | 30.2% |
| No clean resolution | 78 | 1.6% |
Continuation beats reversal better than two to one. Most of the time, a wick beyond a swing level is not a trap being sprung — it is the level genuinely breaking, with the close-back-inside being noise on the way through. Fading every sweep is a slow losing strategy.
Depth is the tell — but backwards from the folklore
If sweeps were stop-hunts, deeper grabs should reverse harder. They do the opposite of what most assume, and it's subtle:
Shallow pokes (a few pips past the level) reverse ~29% of the time — they're mostly continuation. Only the deepest sweeps (beyond ~50 pips) reverse close to half the time, and even then it's a coin flip. A deep sweep is more likely genuine exhaustion; a shallow one is usually just the level breaking. Either way, depth alone never gets you above 50% reversal.
Timing changes the odds more than depth
Where the sweep happens matters more than how deep it is:
| Window | Reversal rate |
|---|---|
| London / New York hours (07:00–16:00 UTC) | ~35% |
| Late / thin hours (20:00–21:00 UTC) | 14–18% |
Sweeps during active, liquid hours are far more likely to reject than sweeps in dead hours — where a poke past a level almost always just keeps going. Reversals need real two-sided participation to form.
What this means
We keep the exact reversal-entry filters private, but the transferable lessons are clear and they run against the popular narrative:
- A sweep is continuation until proven otherwise. The base rate is ~68% continuation — start there, not at "reversal."
- Depth cuts against folklore: shallow sweeps mostly continue; only deep, exhausted sweeps approach even odds of reversing.
- Reversals are a liquid-hours phenomenon. In thin sessions, a level that gets swept is usually just gone.
The edge here isn't spotting sweeps — they're everywhere. It's refusing the reflex to fade them, and only taking the small, filtered subset where the odds actually flip.
Studies like this become the filters inside a documented playbook — the research → playbook → backtest → live loop, locked to one instrument at a time. That playbook runs live on Cortiq.