A strategy that only works on the 30-minute chart
Hold the entry instant and the stop distance fixed and an 18x span of bar sizes pays the same, which makes the strategy far more robust than it first appeared.
- Verdict
- Settled by the data
- Decided by
- Confounded comparison
- The number
- 79% spread becomes 9% once controlled
The hypothesis, as pre-registered
A strategy whose result collapses when you change the bar size is fitted to that bar size, and should be distrusted.
What the test found
Measured naively the alarm looks justified: $19,809 a year at five minutes, $96,448 at thirty, $42,489 at ninety — a 79% spread, with paired t between −2.0 and −2.9 at four of the sizes. A significant effect, not a wobble.
It is two parameters moving with the bar size, and they act at opposite ends. The entry rule fills on the first bar at or after 10:00, which means 10:05 on a five-minute chart and 12:00 on a ninety-minute one — and the 11:00 decision time is separately known to be worth roughly nothing where 10:00 is worth +0.21 average R. Fixing the fill at a single wall-clock instant recovers 84 to 93% of the shortfall at the coarse end.
The second is the stop. One ATR of a five-minute chart is 35 points; one ATR of a ninety-minute chart is 122. “One ATR” silently means a different amount of money on every chart. Holding it at a constant number of points recovers a further 43% at the fine end. Together the two account for 90 to 105% of the whole effect.
Controlled, annual return runs between $83,293 and $90,937 across an eighteen-fold span of bar sizes, the largest paired t against thirty minutes is 0.63, and thirty minutes ranks sixth of eight — it is not even the best cell. The flatness is not emptiness: the drift-null z-score on the controlled family reads +3.0 to +3.7 at every timeframe, better than the naive cell that raised the alarm. It reproduces on two other strategies in the same family.
The general form: any timeframe or bar-type comparison that leaves an ATR-scaled bracket floating is a joint test of three things at once, and the bar clock is usually the one not doing the work. What is genuinely fragile here is the entry instant — which is an operational risk in the deployed code, not a research finding.
What would make this work
Build the controlled family before believing a fragility: fix the fill at one wall-clock instant and the stop at one distance in points, then vary the bar size alone. If the spread survives that, the bar clock is real.
Every insight ends with this line: the exact condition that would turn the finding into a tradeable edge. That is what makes each entry a next experiment rather than a full stop.
Screened September 2, 2026
Run the arithmetic on your own numbers with the deflated Sharpe ratio calculator.
Testing something similar?
The screener puts your version of this idea through the same gate stack, on your market, and answers in seconds. Your first verdict is free.
Related insights
An edge that only exists in the last three years
Ninety-eight percent of a six-stream book's lifetime profit sat after 2022, which is why every result is now split by era before it ships.
SettledPicking the strategy with the best return-to-drawdown ratio
Return persists weakly out of sample and the drawdown ratio does not, so rank on return and correlation and let the portfolio shape the drawdown.