forex-scalping.com

The previous day's high and low are only as good as the definition of "day". Two brokers can show different levels on the same pair because their daily candle closes at different times.

Common setups:

• GMT+2/+3 servers: the day closes at 17:00 New York, giving five daily candles per week. This matches how most institutional desks see the day.
• GMT servers: the day closes at midnight London, and a small Sunday candle appears. That Sunday candle distorts Monday's "previous day" levels.
• Charting platforms: TradingView and others may use yet another session definition depending on the data source.

Practical consequences:

• A level that held "perfectly" on one chart may be a few pips off on another.
• Monday's previous-day levels are most affected. Using Friday's high and low is often cleaner.
• If you backtest PDH/PDL setups, use the same day definition as your live platform.

Which daily close does your platform use, and have you noticed level differences between charts?

FXS Prop Desk wrote: If you backtest PDH/PDL setups, use the same day definition as your live platform.

This one matters more than it sounds, because a lot of free historical data is stored in UTC, so if you download it and build daily bars yourself you get a UTC midnight day by default, Sunday stub and all, even if your live platform runs on a New York close. Then the backtest is testing levels you never actually see live. A quick sanity check is to pull a handful of daily highs from your platform and compare them with the ones your backtest generated for the same dates. If they don't match on Mondays and around the clock change weeks, the definitions are out of line. Those weeks are sneaky because US and UK clocks change on different dates, so the 17:00 New York close shifts relative to London for a week or two.

I lost a clean Monday fade last month because my TradingView daily closed at midnight London while the broker daily closed at 17:00 New York. The previous day high I marked was Friday evening Asia, not the New York close high the platform was using for stops.

Fairman is right about UTC downloads. When I rebuilt daily bars from a free tick dump without forcing New York close, Monday PDH was wrong on 11 of 20 sample weeks.
The illustration overlays New York close PDH against a midnight London daily on the same EURUSD week. The gap is often 4 to 9 pips, which is enough to turn a scalp stop into noise.

Does your live server match the platform you mark levels on? I keep a standing 5 pip buffer when they disagree, and Friday high low is my Monday default when a Sunday stub exists.