RESEARCH

Does the opening range breakout work? 4 markets, 16 years

Study details

Measured
2010 - 2026 one-minute bars on four markets, split into 2010-2017, 2018-2021 and 2022-2026
Instruments
Nasdaq futures, gold, DAX and FTSE. Gold, DAX and FTSE are free spot and CFD proxies rather than the futures themselves
Method
for every 30-minute slot of the day, take the high and low of the first 15 minutes, enter on the first one-minute close beyond it, stop at the range width and target twice that. Then take the same trade in the opposite direction and halve the difference. Costs appear identically in both, so they cancel and what is left is the direction information alone
Result
the direction value is inside 0.023R in all 36 cells. Slot rankings do persist, with rank correlations of 0.42 to 0.88, but those rankings correlate 0.87 to 0.97 negatively with the cost each slot pays. Choosing the best hours on the earlier years left the best of the four at 0.004R and the other three clearly negative
Contents
Opening range breakout is the first mechanical strategy most futures traders meet, and it comes with a second claim attached: that each instrument has its own hours, and trading only those hours is where the money is. Both halves are testable on free data. We tested them on four markets and sixteen years, and wrote down what would count as a pass before looking.

1. The tool that makes this cheap to answer

The usual way to ask whether a signal works is to compare it against random entries, which means running hundreds of simulations and arguing about cost assumptions afterwards. There is a shortcut. Take the breakout trade, then take the identical trade with the direction flipped, same moment, same stop distance, same target. Subtract and halve. Commissions, slippage and the spread are in both legs in the same size, so they cancel exactly. What survives is the direction information and nothing else. We call that the direction value. If the breakout knows which way price is going, it is positive. If it does not, it sits at zero no matter how the net result looks.

2. The breakout does not know which way price is going

Four markets, three periods, and the same answer in each. Widening the range to 30 or 60 minutes does not change it: all 36 cells land inside 0.023R, which is nothing. The best excursion before the stop and the worst excursion before the target are the same size, which is what a coin flip produces. That is the first claim gone. Whatever the net result of an opening range strategy is, it is not coming from the breakout predicting direction.

3. The hours really do rank consistently

Here is where it gets interesting, because the second claim appears to pass. Rank the slots by net result in the first eight years, then rank them again in the next four, and the two rankings agree. The correlation runs from 0.42 to 0.88 depending on market and pair of periods, and it is nowhere near chance. A trader who measured the best hours on history and traded only those would, in fact, be picking hours that stay better.
The last column is the whole story. Each slot pays a cost, and because results are measured in units of the stop distance, the same dollar cost is a bigger fraction of R when the opening range is narrow. The slot ranking correlates with the cost ranking at 0.87 to 0.97 negative. The hours that look best are the hours with the widest ranges. They are not better at predicting; they are cheaper to trade as a fraction of what you risk. The persistence is real and it is the persistence of market structure, not of an edge.

4. So picking the best hours does not rescue it

Choosing the three best slots on 2010 to 2017 and trading only those from 2022 onward does improve things, exactly as the cost explanation predicts. It does not make them profitable. Nasdaq futures land at plus 0.004R, which is zero with extra steps. Gold, the DAX and the FTSE stay clearly negative. Optimising the stop multiple and reward ratio per slot, the way the published versions of this do, reduces the loss further and never crosses zero. Walking the selection forward, re-choosing the hours every year on the previous eight, gives the same picture.

5. One cell did survive, and it is worth looking at

We then went back with a second, narrower question: not every half hour of the day, but the specific moments when a market opens or an number is released, and entering on a stop at the range edge rather than waiting for a close beyond it. That is 264 cells. Two of them held up across all three periods, and they are the same setup: gold, the 15-minute range starting at the London open, entered on a stop.
Three periods, the same sign in each, each one significant on its own, and it works in both directions rather than only long. Neighbouring half hours are positive too but only this one holds up in both later periods, which is what you would expect if the London open itself is the cause. And it is still not tradable. The 15-minute range at that hour has a median width of $1.87, so the cost of a round trip is somewhere between 0.10R and 0.25R. The direction value is 0.074R out of sample. The signal is real and it is smaller than the toll. That is a more useful outcome than a flat zero, because it shows what the search is actually limited by. There is a direction effect at the London open in gold. It is simply not big enough relative to the width of the range it is measured against.

6. What this does and does not settle

It settles that the breakout carries no direction information on these four markets, and that the time-of-day ranking people fit is a cost artefact. Those two results are robust to range length, stop multiple and reward ratio, and the first one does not depend on the cost assumption at all. It does not settle whether some particular published variant is profitable, because the rules that matter most for a result like that are the filters and exits, and those are usually not published. What we can say is that if such a version works, the working part is not the breakout and it is not the hour. Limitations Gold, the DAX and the FTSE are free bid-side spot and CFD series, not the futures contracts, and they carry no volume. Only the Nasdaq series is an actual futures contract. The conclusions that rest on the direction value hold regardless, since it cancels cost, but the net results in section 4 depend on cost assumptions we had to make. The second stage was designed after seeing the first stage, so the gold result carries some search bias even though its three periods were split in advance. It is reported as an observation, not as a validated finding. Entry, stop placement, the 240-minute time exit and the cost figures are our assumptions, not the rules of any particular published strategy. A different set could produce different net numbers, though the direction value is insensitive to all of them. The sample ends in September 2026 and begins in 2010, so it covers one broadly rising regime in equities and one in gold.

Related reading

← All research

Originally published by TraderWe on October 4, 2026. You may quote and link to this page. Republishing the full text without a link back to the original is not permitted.

0 replies

Sign in to reply →