How Often Is the Previous Day's High or Low Taken? Measured

The short answer

The previous day's high and low, PDH and PDL, are the two levels almost every intraday trader marks. We measured what happened to them on every weekday for gold, silver, oil, the Nasdaq, the Dow, the S&P 500, the DAX, four forex pairs, Bitcoin and Ethereum: 13,478 days, 1 Jul 2022 – 18 Sept 2026. On 95% of them the day opened inside yesterday's range, and every figure below is about those days.

At least one of the two was taken by 17:00 New York on 85.0% of days. Both went on only 11%. Which one goes is not a coin flip: the side nearer the open went first about four times in five, and a day that opened in the upper half of yesterday's range took PDH 70% of the time and PDL 27%.

What happens after is the part most plans get wrong. Once one side was taken, price went on to take the other on only 13% of days, and the day closed past the side it had taken half the time (50%). The textbook PDH/PDL sweep trade, taken 8,954 times, averaged +0.03R a trade before costs. The levels tell you where price is likely to go; they do not tell you that it will turn there.

How often PDH and PDL are taken

A side counts as taken when any 15-minute bar trades beyond it before 17:00 New York. Across all 13 markets PDH went on 51% of days and PDL on 45%; the high's small lead most likely reflects four years of rising gold, indices and crypto. Neither was taken on 15% of days: an inside day.

The markets are strikingly alike. Every one of the eleven that trade a session took a side on 83–88% of days. Bitcoin and Ethereum sit lower (80%), and not because of the weekend: their Mondays took a side as often as any other day.

Which side goes first: the open decides

Where the day opens inside yesterday's range says more about which side goes than anything else we measured. Price usually reaches the level it is closest to first, and often that is the only one it reaches.

Both rows point the same way, and they are not independent. The next day opens where yesterday closed, and an up day closes above where it opened, so it tends to hand the next day an open nearer PDH. The open's position is the sharper of the two, so if you know where the day opened, yesterday's candle probably adds little. And the far side is not out of reach: from the upper half, PDL still went on 27% of days.

When it happens

A new trading day starts at 18:00 New York, so the Asian session gets the first look at yesterday's levels. One side had gone by midnight on 38% of days, by 02:00 on 42%, and by 12:00 on 80%. The first take came in the Asian evening on 42% of days, in London on 18% and in New York on 25%.

Each market takes its levels in its own session. USD/JPY and AUD/USD, which trade actively in Asia, had taken one before London on 55–56% of days. The US indices had on 30–37%: they tend to wait for the cash open. The DAX jumps from 32% to 73% across the Frankfurt morning. If a level is still untouched when your market's main session opens, that session is where it usually goes.

After the first side: continuation, not reversal

A popular reading of PDH and PDL is that a sweep of one is a trap that sends price to the other. The measured answer: once one side was taken, the other went too on 13% of days. Price ran at least a quarter of yesterday's range past the first side on 64% of days, and half a range on 40%. Where the day finally closed, at 17:00 New York:

The same shape on every market: about half the days closed beyond the level they took, most of the rest closed back inside, and a full reversal through the other side was the rare case at 6.5%. “Back inside” at 43% is the number that keeps the sweep idea alive, and it is the one the trade below puts to the test.

The PDH / PDL sweep trade

We fixed the trade before looking at any result, in the form it is usually taught as a liquidity sweep. After price takes one side of yesterday's range, wait for the first 15-minute candle to close back inside it. Enter at that close, as long as it comes before 12:00 New York and before the other side has been taken. The stop goes at the furthest price the sweep reached; the target is yesterday's other extreme, or its 50%. Anything still open at 17:00 is closed there. The setup triggered on 70% of days.

The far-side trade won 8.2% of the time, which looks hopeless until you see the payoff. The stop sits just past a sweep that is usually shallow, while the target is the whole of yesterday's range away, so a win paid a median 8.1 : 1. The two cancel almost exactly: +0.03R a trade before costs, where a price moving at random would score zero. Aiming at the 50% won 22% of the time at 3.9 : 1 and came to +0.01R. A nearer target buys a higher win rate with a smaller payoff, and nothing else.

Of the seven filters, swept in the Asian evening (18–02) came out best at +0.10R ± 0.10. Looking at seven filters, one landing about two standard errors from zero is roughly what chance alone produces, so treat it as a hypothesis to test in your own replay, not an edge to trade. The ± column is the uncertainty, two standard errors: a figure inside it cannot be told apart from zero.

Every market

The same far-side sweep trade, market by market.

No market showed an expectancy clearly above zero. Notice the width of the uncertainty: about ±0.17R on gold from 733 trades. A trade that wins about one time in 12 needs hundreds of samples before its result means anything, which is why a week of screenshots settles nothing. The sample size calculator shows how many trades you need.

What this means if you trade PDH and PDL

• Expect the nearer level to go. From the upper half of yesterday's range, PDH went 70% of the time. That is a target worth planning around, not a wall.
• Don't expect the other side next. After one side was taken, the other followed on 13% of days. A sweep of PDH is not, by itself, a reason to aim at PDL.
• Know your market's clock. Asia takes the levels on the yen and the Aussie; the cash open takes them on the indices. An untouched level at your session's open is live.
• Test your version, not the textbook one. The bare sweep-and-fade is worth about zero. Any edge lives in what you add: a shift in structure after the sweep, a fair value gap entry (measured here), a session window. Measure each against this baseline.

That is what the key levels backtester is for: PDH, PDL and the day's opens are drawn on the chart as each day forms, and you replay it one candle at a time with the future hidden, apply your own rules and log the results. The Asian range statistics and CRT statistics run the same test on the session and candle-range versions of the idea.

How we measured it

CRTLAB's own 15-minute price history, 1 Jul 2022 – 18 Sept 2026, weekdays only, for every market except the DAX (from 11 Mar 2025) and crypto (to 21 Sept 2026). A trading day runs from 18:00 to 17:00 New York, the CME day that the CRTLAB daily candle uses, and the 17:00 settlement hour belongs to neither day. PDH and PDL are the highest high and lowest low of the previous day's bars; Monday's previous day is Friday. Days that opened outside yesterday's range (5.1% of all days, most often on the DAX and crypto) are counted but left out of the percentages, because one side is taken before the day begins. A 15-minute bar that took both sides at once (0.2% of days) counts as both. The range sizes are the last year only; every percentage uses all four years.

The trade is walked forward on 15-minute bars. A bar that touched both the stop and the target is scored as a loss, because we cannot see inside it, which is the cautious choice. No spread or commission is included: this measures the market's behaviour, not a broker. The filters were chosen before the results were seen.

Frequently asked questions