A candle shows open, high, low, close. The one thing it doesn't say is which came first, the high or the low. That ordering determines whether a long trade was stopped out before it hit its target, or after.
A common simplification assumes the path goes open, then high, then low, then close for a down bar, and open, low, high, close for an up bar. It's a heuristic and often wrong. Real paths wiggle: up, down, up, down, with the extremes reached at arbitrary moments.
For strategies using wide stops and targets relative to bar range, this barely matters. For anything tight, it's the whole game.
Researchers sometimes try to reconstruct likely paths using lower-timeframe data, so that one-minute bars are tested using their constituent one-second bars, and so on. The method works if the lower data exists. It also depends on sample quality.
The honest approach is to admit the limitation. A strategy whose result depends on the intrabar order isn't validated by bar data, and the only fix is better data.
What share of trades flip result depending on which came first inside a bar, the high or the low?
High and low order inside a bar: the part nobody recorded
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
-
FXS London Scalper
- Posts: 979
- Joined: Sun Oct 18, 2026 12:25 pm
Re: High and low order inside a bar: the part nobody recorded
The high/low ordering issue is one of the most common sources of backtest optimism. Using M1 data within H1 bars, or tick data where available, solves most of it. Checking how many trades had both the stop and target inside one bar is a quick way to see if it matters.
FXS Official Team | London Session Desk
-
Amanhasnoname
- Posts: 2060
- Joined: Fri Oct 30, 2026 9:45 am
Re: High and low order inside a bar: the part nobody recorded
Last spring I rebuilt a pullback scalp on M1 EURUSD and checked every trade where both stop and target sat inside the same bar. Out of 214 entries across eight weeks, 37 had both levels inside one candle. Of those 37, reversing the assumed high-low path flipped the result on 19 of them, which is roughly half of the ambiguous set and about nine percent of the whole sample.
I stopped trusting OHLC-only backtests for anything with a four-pip stop after that. Now I replay with tick files for the pairs I actually trade, and I flag any trade where stop and target touch the same bar as unresolved rather than counting it as a win. That cut the paper expectancy by almost a pip per trade on that system, but live fills lined up closer.
The heuristic of open-high-low-close on a down bar was wrong more often than I expected once London opened. Quiet Asian hours matched it better. For tight scalps the share that flips is high enough that bar data alone is marketing, not proof.
I stopped trusting OHLC-only backtests for anything with a four-pip stop after that. Now I replay with tick files for the pairs I actually trade, and I flag any trade where stop and target touch the same bar as unresolved rather than counting it as a win. That cut the paper expectancy by almost a pip per trade on that system, but live fills lined up closer.
The heuristic of open-high-low-close on a down bar was wrong more often than I expected once London opened. Quiet Asian hours matched it better. For tight scalps the share that flips is high enough that bar data alone is marketing, not proof.