Opening Range Breakout Strategy: What 16 Years of ES Data Show
The opening range breakout is one of the most-quoted ideas in day trading and one of the easiest to blur with other kinds of range break. This page separates the experiments, publishes their exact questions and shows what 16 years of ES five-minute data do — and do not — establish.
Quick answer
Does the opening range breakout strategy work? The measured answer is narrower than a yes or no. The opening session carries useful structural information, but these statistics do not test a complete entry, stop, exit, fee and slippage rule. They are not a strategy win rate or a claim of profitability.
Two different breakout tests appear below. The archived 83.5% result concerns a fixed 18-bar opening range and asks whether the opposite session extreme survives to the close. The 52.1%, 62.6% and 70.6% results concern a rolling 10-bar range and ask whether price closes back through the breakout level within 5, 10 or 20 bars. They use different ranges, outcomes, horizons and samples; they are not the same experiment or complements.
What is an opening range breakout strategy?
An opening range breakout, or ORB, begins with the high and low of a fixed period after the session opens. A break is recorded when price moves beyond that band. Traders commonly choose 5, 15 or 30 minutes, but the archived test on this page used the first 18 five-minute bars — 90 minutes — and the first later close outside that range. Toby Crabel popularised the setup; Al Brooks placed it inside a broader bar-by-bar framework about when session extremes form.
The two breakout tests on this page
| Question | Archived opening-range test | Current rolling-range test |
|---|---|---|
| Range | First 18 five-minute bars | Immediately preceding 10 bars, no wider than 0.5× ADR5 |
| Event | First later close outside the opening range | First qualifying closing break per window and day |
| Outcome | Does the pre-breakout opposite session extreme survive? | Does a later close cross back within 5, 10 or 20 bars? |
What does the opening session reveal?
The first study asks only whether one side of the day’s final high–low pair has already printed. It does not choose a trade or measure profit.
| Checkpoint | Brooks framework | Measured ES | Sample |
|---|---|---|---|
| After the first 5-minute bar | ~20% | 26.3% | 3,946 sessions |
| 35 minutes into the session | ~50% | 58.3% | 3,946 sessions |
| 90 minutes into the session | ~90% | 81.2% | 3,946 sessions |
By 35 minutes, one side of the final session range had already printed in 58.3% of the 3,946 archived sessions. By 90 minutes that figure was 81.2%. Those are timing base rates, not directional forecasts.
How often did the archived 18-bar ORB preserve the opposite extreme?
| Claim | Brooks framework | Measured ES |
|---|---|---|
| Bull days: session low forms in the first third | ~90% | 83.3% |
| Bear days: session high forms in the first third | ~90% | 80.4% |
| 18-bar opening-range break: opposite session extreme holds | ~90% | 83.5% |
The archived scorecard reports 83.5% across 3,578 valid sessions. Its original executable scorer was not retained. A later reconstruction used the first 18 bars as the opening range, took the first later close outside it, and asked whether any later wick exceeded the opposite session extreme already in place before the break. That reconstruction returned 83.7% across 3,574 sessions — close, but not identical.
We therefore keep 83.5% labelled as the archived result rather than present the reconstruction as its exact source. The outcome does not mean price never re-entered the opening range, and it is not a strategy win rate.
How often does the current rolling-range event fail?
MADDOG’s live event uses the high–low of the immediately preceding 10 five-minute bars. That range must be no wider than 0.5 times ADR5, the mean daily range from available prior complete RTH sessions up to five days. The event is the first qualifying bar in the window whose close finishes above the range high or below the range low. A failure is a later close back through that breakout level; only the first event per window and day is counted.
| Breakout closes back through its level | Failure rate | Eligible events |
|---|---|---|
| Within 5 bars (25 minutes) | 52.1% | 3,925 |
| Within 10 bars (50 minutes) | 62.6% | 3,922 |
| Within 20 bars (100 minutes) | 70.6% | 3,915 |
ES five-minute bars, 2010–2026, ADR5 production definition. These are the frozen RTH reference tuples used by the current API.
The practical statement is precise: roughly half of qualifying events closed back through their level within 25 minutes, and 70.6% did so within 100 minutes. That does not make 70.6% the loss rate of an ORB trading system; no common entry, stop or exit was tested here.
Is the ORB strategy profitable?
This study cannot answer that question. Profitability requires a fully specified range window, entry trigger, stop, exit, position sizing, fees and slippage. The archived experiment measures whether a session extreme survives; the live rolling experiment measures a close back through one level. Neither is a P&L backtest.
The page also does not rank 5-, 15- and 30-minute ORBs. Those are separate strategy definitions and need a pre-declared comparison with occurrence rates, outcomes, uncertainty and trading costs.
Methodology, provenance and limits
The opening-timing figures come from a frozen scorecard over ES five-minute sessions from 2010–2026. The original scorer was not retained; a later reconstruction reproduced the timing results exactly and the archived 18-bar ORB result within 0.2 percentage points. The rolling-range definition was recovered after its original rates were already known, then frozen before it was carried into other session windows. Its RTH aggregate is therefore a reference rate, not an independent replication.
Separately, the rolling event’s conditional model was specified before its original zero-shot NQ test. The later production ADR5 port was checked on NQ data from 2023 onward: AUC 0.646, calibration error 5.1%, n = 2,637. That validation tests whether the model distinguishes and calibrates rolling-range outcomes; it does not validate the aggregate reference rates or the archived 18-bar ORB result.
Read the full methodology and provenance record, or download the machine-readable appendix.
What does MADDOG measure live?
The live product emits the rolling 10-bar event above, not the archived 18-bar opening-range experiment. It returns the first qualifying event in the selected window, the matching historical reference rates and a calibrated estimate of failure within 10 bars. Definitions and field semantics are in the API reference.
Read real sample emails, or leave an email and one ticker on the homepage and get one read free — no account, no card.
Related measurements
For the complete event definition and all three follow-up horizons, read the dedicated opening range breakout failure-rate study. For the 5-, 35- and 90-minute timing checkpoints and their hindsight limits, read when the high or low of the day forms.