Opening range breakout with an 81% win rate: does it work?
A viral short-form video says a 15-minute opening range breakout wins 81% of the time and is profitable: mark the first two 15-minute candles after the open, enter when a candle closes outside, stop at the other end of the range, take 10 points. A second clip adds a 1:1 version, and both call it a variant of the "School Run".
The exact rules we tested
- Range: high and low of the first two 15-minute candles after the cash open (09:30 to 10:00 New York time; 09:00 Berlin time for the DAX).
- Entry: the close of the first 15-minute candle that closes above the range high (long) or below the range low (short). One trade a day, none after 15:30.
- Stop: the opposite end of the range. Target: 10 points (10 dollars on TSLA). Anything still open is closed at the cash close.
- Costs: a tick of slippage on entry and target, two on the stop, plus commission. A trade counts as a win only if it makes money after costs.
- Calendar: this is a cash-open strategy, so US exchange holidays and Xetra holidays are skipped and US early closes (13:00) are respected. The Xetra holiday dates follow the standard published pattern and were not re-verified year by year, and the TSLA feed has no holiday bars. Runs on the raw CFD clock (every weekday) are kept in the research files and differ only slightly.
- Data and splits: Dukascopy 5-minute bid CFD prices (NQ, ES, TSLA, DAX), 2020-01-01 to 2026-09-30, after costs and slippage; cash-exchange calendar; dev 2020-2023, test 2024-2026-09-30.
Results
| Instrument | Trades | Win rate (95% CI) | Target hit | Expectancy per trade, net (all) | Expectancy, net (test; 95% block-bootstrap CI) |
|---|---|---|---|---|---|
| NQ, 10 pt target | 1,641 | 90.4% (88.8 to 91.8) | 90.4% | -0.84 pts | -1.15 pts (-4.24 to +1.73) |
| ES, 10 pt target | 1,656 | 65.9% (63.5 to 68.2) | 63.2% | -0.27 pts | -0.72 pts (-2.11 to +0.66) |
| TSLA, $10 target | 1,484 | 56.3% (53.7 to 58.8) | 25.3% | $+0.84 | $+0.34 ($-0.06 to $+0.74) |
The tested reconstructions did not pass the pre-registered bar on these data and assumptions (0 of 17 cells). Test-split expectancy intervals include zero for all three. 0 of 17 variant and instrument cells passed the pre-registered bar; the smallest Benjamini-Hochberg q was 0.257.
What the clip says, and what we found
On NQ the win rate was 90.4% and its confidence interval (88.8 to 91.8) lies above 81%, so the headline win rate is reproduced there. On ES it was 65.9% and on TSLA 56.3%, both below 81%, and TSLA is the stock the clip's own backtest screenshot shows (94/116 = 81.03% (not reproduced)). In the clip's own window, January to June 2026, TSLA won 55.9% of 102 trades (interval 45.7 to 65.7), and only 13.7% of trades reached the $10 target. We could not reproduce the screenshot.
Why a 90% win rate can still lose money
On NQ the mean distance from entry to stop is 137.8 points (median 120.7), about 14 times the 10-point target. That is not the width of the opening range: the first 30-minute range averages 114.1 points (median 100.0). The stop is bigger than the target on 100% of trades. A price that drifts without direction reaches the 10-point target before the stop about 91.7% of the time, using the average of each trade's own ratio (R / (R + 10), where R is that trade's stop distance); the ratio of the averages gives 93.2%. Both are idealised: no costs, no time limit, not a calibrated probability for a real session. The strategy hit its target 90.4% of the time. The average win was 9.52 points and the average loss 98.12, so break-even needs about 91.2% wins, and the result is -0.84 points a trade after costs. A high win rate is what a wide stop produces on its own.
The 1:1 version and the "80% to 40%" exit rule
The second clip says the win rate falls from 80% to 40% if you enter on a candle that does not close outside the range, so you should exit. Read literally, the rule can never fire when you enter on the close. Trading it the way the clip's chart does, entering on the touch of the range edge, the win rate was 65.7%, 63.3% and 62.6% when the entry candle closed outside (NQ, ES, TSLA; interval upper bounds 68.8, 66.4 and 65.9) and 40.6%, 35.9% and 43.2% when it closed inside. The lower figure is close to 40%, the higher one is not near 80%. Following the exit advice did not beat staying in the trade on NQ (+3.27 against +4.91 points a trade) or TSLA, and made little difference on ES. This is a descriptive split made after the entry candle closes, not a rule that could be known at entry.
The School Run variant
We also tested our best reading of the original School Run: trade a break of the second 15-minute candle only, stop at the other end of that candle, no target, flat at the close, in two versions (H and H1R). H and H1R are hypothesized second-candle breakout reconstructions; the stop and exit rules are unverified, so no verdict applies to Hougaard's actual method. Win rates were 30.9% to 42.6% for H and 50.0% to 54.6% for H1R, and no test-split t-statistic reached the pre-set bar of 2.
| Cell | Trades | Win rate | Expectancy, net (all) | Expectancy, net (test) | Test t | BH q | Random-direction percentile (test) |
|---|---|---|---|---|---|---|---|
| NQ, H | 1,654 | 37.7% | +6.22 | +3.67 | 0.7 | 0.45 | 97.2th |
| NQ, H1R | 1,654 | 53.4% | +2.92 | +5.57 | 1.4 | 0.26 | 96.0th |
| ES, H | 1,640 | 33.9% | +0.67 | +0.66 | 0.6 | 0.45 | 96.3th |
| ES, H1R | 1,640 | 50.4% | -0.35 | +0.29 | 0.4 | 0.53 | 93.3th |
| DAX, H | 1,641 | 30.9% | +4.97 | +4.97 | 1.2 | 0.33 | 51.3th |
| DAX, H1R | 1,641 | 50.0% | -0.94 | -0.20 | -0.1 | 0.69 | 71.5th |
| TSLA, H | 1,657 | 42.6% | +1.31 | +0.33 | 1.5 | 0.26 | 98.9th |
| TSLA, H1R | 1,657 | 54.6% | +0.75 | +0.29 | 1.8 | 0.26 | 98.0th |
Controls
Each cell was compared with the same trade using a coin-flip direction (1,000 random seeds), a random entry time on the same day, and always buying or always selling after the range forms. On NQ with the 10 point target, the real result sat at the 86.7th percentile of the random-direction control on the test split, below the 95th percentile bar, and the random-time control put it at the 61.8th, so a high win rate shows up even with no skill in picking direction. 0 of 17 variant and instrument cells passed the pre-registered bar (positive on dev and test, test t-statistic of at least 2, above the random controls, still positive with double slippage). Inference: Benjamini-Hochberg on one-sided t-test p (test mean net P&L > 0); t is not an exact test under serial dependence, month-block bootstrap CIs are given alongside. The closest were TSLA cells, positive in both splits, with test t-statistics of 1.5 to 1.8 against the 2.0 required. TSLA's dollar edge per trade by year (literal prices) was 2020: $1.65, 2021: $1.80, 2022: $1.03, 2023: $0.20, 2024: $0.17, 2025: $0.34, 2026: $0.57; no monotonic decline in either series; do not say the edge shrinks every year.
Limits
A second run (v2) was made after an independent audit of the first and fixed several execution and labelling defects. It changed some counts slightly and changed no verdict. The index data are Dukascopy CFD bid prices, not futures, with no usable volume, so fills differ from a futures account. TSLA prices in our cache are not split-adjusted, so a split-adjusted view is reported separately and the two are not comparable year by year. We examined 17 cells and report Benjamini-Hochberg q values alongside the pre-set bar. Bars are 5 minutes, so a touch entry fills at the range level. Days with an incomplete opening range were skipped (8 on NQ, 14 on ES, 8 on TSLA, 59 on the DAX) and exchange holidays were removed as described above. H and H1R are hypothesized second-candle breakout reconstructions; the stop and exit rules are unverified, so no verdict applies to Hougaard's actual method. A confidence interval that contains 81% is compatible with the claim; one that lies entirely above or below it says so. A high win rate here comes from a stop about 14 times the target, not from edge. None of this says a breakout can never work, only that these reconstructions, on this data and these assumptions, did not pass a bar set before we looked.
General information only. It doesn't consider your objectives, finances or needs. Tickfloor holds no financial services licence and never places trades.
Does the opening range breakout strategy really win 81% of the time?
On NQ with a 10 point target and the stop at the other end of a 30-minute range, it won 90.4% of 1,641 trades, but lost 0.84 points a trade after costs. On ES it won 65.9%, and on TSLA 56.3%. The high NQ figure appears where the stop is about 14 times the target.
Why does a high win rate lose money here?
The average NQ stop distance was 137.8 points against a 10 point target, so most trades reach the target. The average win was 9.52 points and the average loss 98.12, so break-even needs about 91.2% wins and the strategy got 90.4%.
Is the opening range breakout profitable?
The tested reconstructions did not pass the pre-registered bar on these data and assumptions (0 of 17 cells). NQ and ES lost money after modelled costs on the 10 point version.
Is this financial advice?
No. This measures a publicly claimed trading rule, not a recommendation. General information only, not personal advice.
How we set the rules, costs and pass bar before running anything: the Tickfloor testing standard. The wider record is on the evidence page.