What Is the Opening Range Breakout? The Rule and What Testing Found

An opening range breakout is an entry rule with no forecast inside it. It marks a level above where the session opened and another below it, then sits out until price reaches one of them. Two freely available studies have run that rule on futures data and reached opposite answers on different contracts: statistically significant average returns on U.S. crude oil futures from 1983 to 2011, and on Micro E-mini Nasdaq 100 futures not one of the five variants tested cleared the author's bar.

Neither result tells you to trade it. What they are good for is showing which assumptions the rule is carrying, and which of those the evidence turns out to be sensitive to.

What is the opening range breakout?

The opening range breakout, usually shortened to ORB, turns the start of the session into two trigger levels — one above, one below. No position is held while price stays between them. If price crosses the upper level from beneath, the rule goes long; if it crosses the lower one from above, the rule goes short. That is the whole mechanism.

The unusual part is that the trader chooses the distance and the market chooses the side. Holmberg, Lönnbark and Lundström — whose 2012 Umeå working paper became a 2013 article in Finance Research Letters — name a 1990 book by T. Crabel as the origin of the rule, and credit him with the principle underneath it, which they call Contraction-Expansion. Its content is a claim about two states rather than two directions: stretches in which the daily price movement stays small give way to stretches in which it is big, and then back again. On that reading the levels are a filter for the second kind of day, and the direction is a by-product of which level happened to be reached.

What has to be true for the rule to pay is narrower than that, and the authors do not leave it implicit. Once price has travelled the chosen distance from the open, continuation has to be better than a coin flip, because the entry is being taken at a worse price than the open. That is a momentum claim, and they anchor it to Jegadeesh and Titman (1993) rather than asserting it as obvious.

How is the opening range actually defined?

There is no single definition, and the two studies here share no parameter. One sets the trigger as a percentage step away from the opening price, with no time window involved in setting it. The other takes the high and the low of the first 25 minutes of the 09:30–16:00 ET session and uses those as the levels. A number produced under one says nothing about the other.

The percentage version is estimated rather than picked. The authors take the mean and the standard deviation of the open-to-close log return, place the trigger at a move a normal distribution would call unlikely, and tune how unlikely by moving the tail probability. A stricter setting is literally a rarer day. The Nasdaq version needs no estimation at all: the levels are wherever the first 25 minutes traded.

Both are the opening range breakout as their own authors define it. Which is why any figure quoted for "the strategy" belongs to one definition, on one contract, over one stretch of years, and travels no further than that.

Does the opening range breakout work on futures?

Neither study covers futures as a class. Between them they cover two contracts, and the answers differ: significant positive average returns on crude oil futures, and no variant clearing the bar on Micro E-mini Nasdaq 100 futures. The crude oil result also arrives with a limit its own authors put on it.

Start with that limit, because it is the more useful half. The paper cuts its sample into three periods and reports that the full-sample finding does not hold up period by period. From March 1983 to June 1992 the long side was insignificant at every threshold reported: 260 trades at the loosest setting returning 0.0334, with a bootstrap p-value of 0.2539. From mid-1992 to October 2001 several settings return nothing worth having: 0.0068 on the long side at one threshold, with a p-value of 0.4546, and negative averages on the short side at three of the five, two of them carrying p-values of 0.6814 and 0.6420. Only the last period, October 2001 to January 2011, is strong, and the authors identify it as both the driver of the full-sample result and the most volatile of the three.

The full-sample numbers it qualifies are real enough. Across 6,976 daily observations, with significance assessed by a bootstrap following Brock, Lakonishok and LeBaron (1992), the long side at the loosest threshold took 738 trades, 60.6% of them positive, for an average of 0.2019 in the paper's percentage units and a p-value below 0.0001. Tightening the threshold lifts both the hit rate and the average — 80 trades, 71.3% positive, 0.4027 — but the trade count falls by about 89% to buy that.

The authors read the rule as directional and, in their words, "basically long volatility", and they relate the time-dependence to Gençay (1998), cited there for the link between mechanical rules and markets that trend or run volatile. The published tests of the head and shoulders pattern make the point from another direction — that pattern paid on two currencies and not on U.S. equities — and a result attached to one sample is not a result about the rule.

Why did every opening range variant fail on Nasdaq futures?

For two different reasons, worth keeping apart. Four of the five variants lost money outright once costs were charged. The fifth made money and still failed, because the study required more than a positive number.

The bar was five conditions holding at once: at least 30 out-of-sample trades, a permutation test under 0.05 where one applied, a positive net return after friction, the same direction of return in every test year, and an out-of-sample t-statistic of at least 2.0. Parameters were fixed on a training window before each test window. The study is an arXiv preprint by an independent researcher, not peer reviewed, and worth weighing accordingly; it ran 14 signal families over 947 trading days of five-minute Micro E-mini Nasdaq 100 data from December 2021 to August 2025. The opening range family among them used the 09:30–09:55 ET window.

Three of the four losers are plain enough: long held one bar, 0.82 points down net over 447 trades; short held one bar, 3.45 points down over 428; short held fifteen bars, 2.16 points down. The survivor was the long side held fifteen bars — 2.82 points net across 447 trades, a 55.5% win rate — but its t-statistic was 1.50 against the 2.0 required, and the year-by-year record ran 1.42 points down in 2022, 2.43 up in 2023 and 7.04 up in 2024. Mesfin identifies that shape, a record resting on a single good year, as the failure mode that recurred most often across all fourteen families.

The fourth deserves separating out. Instead of entering on the break it waits for price to return within five points of the level, with a 20-point stop. Over 83 trades it was stopped out 80.7% of the time and lost 4.44 points net, which the author explains by most apparent MNQ breakouts failing, leaving that entry on the wrong side of the turn.

He is explicit about scope: MNQ may not stand in for other equity index futures, and he will not carry the findings across to commodity or currency contracts until someone has tested them there.

How much do transaction costs change the result?

On the Nasdaq data, costs decide the outcome rather than trim it. Mesfin puts the maximum gross return across all fourteen families at about 0.07 to 1.50 points per trade, against friction of 2.0 points. The ceiling on what the signals could earn therefore sits below what it costs to act on them, which is why he treats the failures as structural rather than unlucky.

He puts the equilibrium gross edge on five-minute bars at one to two points per trade, and reads the shortfall as competitive rather than statistical: a pattern anyone can see in public OHLCV data gets traded against until what is left of it is down around what the trade costs the marginal participant. That ceiling is stated for single-bar directional predictions, not for everything. His own comparison table lists the opening range long at 4.82 points gross, outside that 0.07–1.50 ceiling — a reminder that a preprint is a preprint.

The crude oil study sits at the opposite extreme and says so. It charges nothing: no commission, no spread, and every entry filled at exactly the threshold price. The authors put real costs in that market at 0.04%, or 0.08% round trip, and accept that this eats into the profit. They argue the test understates in the other direction too, since forcing every exit at the daily close keeps the days that reversed intraday, which a stop would have cut short.

The Nasdaq study handles the other half of the execution problem by timestamping. Entries are booked at the open of the bar following the one that produced the signal, never at that bar's own close, so nothing in the test executes at a price the rule could not yet have seen. That gap between a modelled fill and a real one is the same gap that separates a simulated fill from a live one, and it does not narrow because the rule is simple.

When does a funded account close a breakout trade for you?

At 4:45 PM ET, every trading day. Both studies force an exit too — the crude oil test at the daily close, the Nasdaq test after a set number of bars — but on an account it is a rule rather than a modelling choice. On our programs every position must be closed by 4:45 PM ET, and anything still open is closed automatically, with the result counting toward the balance and the drawdown. The trading hours rules set that out.

So a breakout rule that specifies no exit of its own inherits that one. The trading day itself runs on the CME session, opening at 6:00 PM ET and running through to the following afternoon's flat time, which means a position taken on a Sunday evening belongs to Monday, and that nothing can be carried overnight or across a weekend. Both papers treat the exit as part of the rule rather than an afterthought, and here it is fixed before you start.

How does a low win rate interact with the drawdown line?

A breakout rule can lose most of its trades and still be the rule as tested. Across Mesfin's five variants the win rate ranged from 19.3% to 55.5%; Holmberg, Lönnbark and Lundström report proportions of winning trades from roughly 33% to 83%, depending on threshold, sub-period and side. Neither range forecasts anything about your account.

That matters because of how the line moves. On a funded account the drawdown line only rises on a new highest closing balance, never on an unrealised intraday gain. A sequence of failed attempts therefore spends a budget that will not be replenished before the session ends.

Sizing is where the two meet, and neither study helps. Both report results per trade, in percentage returns or in points; neither models position size at all. Turning the distance between an entry and a breakout level into a number of contracts is left entirely to the trader, and on a funded account the budget for that arithmetic is the published distance to the line rather than a figure you pick. That conversion is worked through in futures position sizing.

FAQ

What time is the opening range in futures trading?

It depends whose rule you are reading, and the two studies here do not agree. The Nasdaq micro study uses 09:30–09:55 ET — the first 25 minutes of the 09:30–16:00 ET regular session it defines — and takes the high and low of that window as the levels. Holmberg, Lönnbark and Lundström use no window at all: their trigger is a percentage distance from the opening price, and it fires whenever price reaches it during the day.

Who invented the opening range breakout?

Holmberg, Lönnbark and Lundström name T. Crabel as the originator, citing his 1990 book Day Trading With Short Term Price Patterns and Opening Range Breakout, published by Traders Press. They credit him too with the principle they say it rests on, the Contraction-Expansion principle, under which daily price movement cycles between quiet stretches and violent ones. Framed that way, a breakout level sorts one kind of day from the other rather than predicting direction.

Does the opening range breakout need a stop loss?

The two studies answer differently. Holmberg, Lönnbark and Lundström tested the rule with no stop, no trailing stop and no profit target, closing everything at the daily close, and noted that their test therefore keeps days where the move reversed intraday, days on which they say a stop would cap the loss in practice. Mesfin tested a pullback entry carrying a 20-point stop and recorded an 80.7% stop-out rate on it across 83 trades.

Why do intraday backtests look better than live results?

Mesfin's preprint gives four causes: parameters tuned on the same data used to test them, fills priced at mid-market instead of at the bid or the ask, friction left out altogether, and a literature in which the attempts that failed are never written up. His answers were a friction charge held fixed throughout, parameters set on the training window only, and every failed signal kept on the record instead of being retuned. Most people who attempt day trading do not make money; see our risk disclosure.