Home › Learn › Lesson 26

Two-candle reversals — engulfing, tweezers, dark cloud cover, piercing line

By TradingPrimer · Published 3 Sep 2026

Two-candle reversals are usually taught as a small zoo: four names, four little drawings, four descriptions of who overwhelmed whom. This lesson does something different. It shows that the four are two thresholds on a single axis, that one of the two classical definitions cannot physically occur on a crypto chart, and that the entire bullish engulfing rule collapses in a 24/7 market to one comparison you can make in half a second. Then it prices the thing, because that is where the surprise is: the version of the pattern every guide calls strongest is the one that hands you the worst trade, and the gap between them is measurable to the penny.

Illustrative example with hypothetical numbers, drawn to the price scale on the right. The full arithmetic is in the table under “Why does the deeper close pay less?”

What is a two-candle reversal actually claiming?

That one side was in control, the other side arrived, and the second candle undid a measurable share of what the first one achieved. Every one of the four names in this lesson is a claim about how much got undone.

Lesson 24 read single candles as a record of a fight inside one session. A two-candle pattern is the same idea across a boundary: candle 1 shows a side winning, candle 2 shows it losing ground back. And that raises the question the pretty drawings skip — losing back how much, and measured against what?

Before any of that, though, there is a precondition our course material states first and most guides mention last. The checklist for the engulfing pair opens with a flat requirement: the market must be trending. Not “preferably”. First line.

The reason is not aesthetic. Our own course notes put it in terms of force: a pattern is only worth anything when there is something for it to reverse, and the force pushing back has to be big enough to prove itself against the force that was already there. If nobody was in control, nobody can be overwhelmed. A green candle that swallows a red candle in the middle of a quiet range is a fact about two candles and a claim about nothing.

That precondition also has a price, which is easier to feel than the argument. Hold everything else in this lesson constant and put the identical pattern inside a 3,000-point range instead of a trend, so the next level is only 3,000 points away rather than 6,000: the trade goes from 1.60R to 0.40R, and the win rate you need to break even climbs from 38.46% to 71.43%. Same two candles. The only thing that changed is how much room the market left you, and diagnosing that first is what the precondition is really asking for.

One more boundary before the mechanics, because it is the honest frame for the whole lesson. Our course notes list the habits that empty accounts, and one of them is named plainly: reading the signal off one to three candles instead of off the whole move. A two-candle pattern is exactly one to three candles. So treat what follows as a timing device inside a read you already had — never as the reason for the trade. Everything in this lesson makes that distinction cheaper to hold, because once you can price the pattern in four seconds, most of them disqualify themselves before you have time to fall in love.

Why are engulfing, piercing line and dark cloud cover really one pattern?

Because all three are defined by the same measurement, differing only in where they put the threshold and which direction they point. Write it down once and the zoo becomes an axis.

Take candle 1 with an open o₁ and a close c₁, so its body is b = |o₁ − c₁|. Now measure how far candle 2 closed past candle 1's close, in the recovering direction, and call that x. The penetration is p = x / b. That single number gives you all four names:

Four of the five entries are two thresholds on one axis in two directions. The tweezer is the genuine outsider, and that turns out to matter enormously.

The gap that cannot happen

Now the part that changes what you can and cannot trade. The classical definitions of the two half-penetration patterns both require a gap. A textbook dark cloud cover has candle 2 opening above candle 1's high — that is what makes it a trap, since the market first sees a bullish jump and only then gets sold into. A textbook piercing line mirrors it: candle 2 opens below candle 1's low.

A gap needs a market that stops trading. Crypto spot and perpetual pairs do not stop. Each candle's open is the first trade of the interval and candle 1's close was the last trade of the one before, so the two are the same print, or a tick apart. In symbols, o₂ ≈ c₁, always.

So we counted. In a driftless random walk of 400,000 contiguous candles — 399,974 adjacent pairs — the number of times candle 2 opened above candle 1's high or below its low was zero. That is not a statistical result, it is an arithmetical one: on a continuous series the open cannot escape the previous candle's range because it is the previous candle's last price. The classical dark cloud cover and the classical piercing line do not occur in continuous crypto at all.

Which makes something in our own course material look much smarter than it first appears. The prose describes the gap, as any candle book does. But the operative checklist — the one you actually trade — drops it entirely, and keeps only: sustained trend, opposite-coloured candle, and close at least 50% into the previous body. That is not a simplification. It is the only version of the pattern that can exist on a chart that never closes.

And the engulfing shrinks to one comparison

The same contiguity does something even more useful to the engulfing rule. The full definition has two conditions: candle 2 must open at or below the bottom of candle 1's body, and must close covering the whole body. But if candle 2 opens exactly where candle 1 closed — the bottom of a red body — the first condition is satisfied automatically, every single time.

What is left is a single test:

Note also what does not need to be covered. Our course material is explicit that candle 2 must cover candle 1's body and “need not cover the wicks”. People routinely wait for the wicks to be swallowed too, which is a stricter pattern with a different name and, as the next sections show, a considerably worse price.

Reducing a rule to one comparison is normally good news. Here it is a warning, and the next section explains why.

How often does an engulfing appear when nothing is happening?

Far too often to be a signal by itself. One comparison is a low bar, and a low bar gets cleared by accident.

To put a number on the accident rate we need a market with nothing in it. So, using the same method and the same seed as Lesson 25 so the two lessons' figures are comparable: a driftless random walk of 400,000 candles, each built from 60 sub-steps so the open, high, low and close come from a real price path rather than being invented, volatility 0.4% per candle, seed 20260903. No trend, no news, no participants, no memory. Whatever turns up is coincidence. That is the point — it is the floor.

An engulfing pattern in one direction or the other shows up on a quarter of all candle pairs by pure chance. Translate that into a trading day and the problem becomes obvious:

Trends do not turn twenty-four times a day. On a 15-minute chart the genuine turning points are buried under shapes that mean nothing, and no amount of care in reading them helps, because there is nothing in them to read.

This is also a quiet vindication of a line in our course material that otherwise sounds like folklore: this pair works well on the 1H, 4H and 1D frames. Now it has a mechanism. On the daily, the pattern arrives about once a week per direction — the same order of magnitude as the event it claims to announce. On the 15-minute it arrives two hundred times more often than trends reverse. The recommendation is not about the candles being “more reliable” higher up. It is about supply.

Two things to be careful about with that table. It is a frequency, not a win rate — nobody is right or wrong 12.521% of the time. And it is a floor, not a forecast: real markets trend, and trending markets produce these shapes at a different rate. What the floor establishes is that a shape appearing on one pair in eight cannot, on its own, be evidence that somebody meant something.

Where do the entry, the stop and the target actually go?

Entry on the candle after the pattern completes, stop beyond the extreme wick of whichever of the two candles reaches further, target at the next structural level on the frame you are trading. That is exactly how our course material teaches it, and the second one is where the money is.

Look at where those two lines sit relative to each other. The stop is not a decision you made about how much you are willing to lose. It is wherever the pattern happened to reach, and the entry is wherever candle 2 happened to close. Both are handed to you by two candles. That is the mechanism the rest of this lesson is about, and it deserves its own notation.

Set up the structure we will use for every calculation from here. All the numbers are illustrative, chosen so you can check each step by hand:

• Candle 1 is red: opens 64,000, closes 62,000. Its body is b = 2,000.
• The lower of the two lows — where the stop goes — is 61,500. So the distance from candle 1's close down to the stop is d = 500.
• The next resistance above is 68,000. From candle 1's close that is D = 6,000.
• Candle 2 closes somewhere above 62,000. Call that distance x.

Now the trade prices itself, and the algebra is two lines:

This is the same mechanism Lesson 24 found for single candles, where R:R = D ÷ h − 1 with h the pattern's height. Written in x it becomes sharper, because x is the quantity the pattern's name is graded on — and look at what is missing from those two lines. b does not appear. Candle 1's body height, the denominator of every textbook definition in this lesson, has no effect on your risk, your reward or your reward-to-risk.

Before we spend that observation, one measurement from the simulation, because it changes how the stop feels. Across every bullish engulfing in the data, the median value of x ÷ (x + d) was 70.2%. On a typical engulfing, roughly seven-tenths of your stop is not protecting the level underneath the pattern — it is paying for the distance candle 2 travelled before it closed. The part of the stop that does the job the stop is for is the smaller part.

There is a second cost hiding in the same place, and it belongs to a different lesson. The stop is the extreme of a two-candle window — the smallest structure a chart can offer — while the target is a level from a much larger frame. Our course notes on moving a stop are blunt about this: a stop has to correspond to the structure of the frame you are actually trading. So the textbook engulfing trade is a frame-mixed trade by construction, and Lesson 21 already priced what mixing frames does to the reward. Worth knowing you are doing it.

Why does the deeper, more convincing close pay less?

Because the thing that earns the stronger name is the thing that moves your entry away from your stop and towards your target at the same time. One distance, two penalties.

Run the master table. Same candle 1, same stop at 61,500, same target at 68,000 — the only thing changing is where candle 2 closed:

The ranking runs backwards. Every guide orders these as piercing line < engulfing < huge engulfing, weakest to strongest. The prices run 3.33R > 1.60R > 0.44R, best to worst. Both orderings are correct. They are simply measuring different things, and only one of them is measuring your account.

The name divides by the wrong thing

Here is the cleanest way to see it. The pattern's grade is x ÷ b. The trade's quality is x compared with d and D. Two different denominators, and neither one knows about the other.

So hold the pattern fixed at exactly a minimum engulfing — p = 1.00 every time, textbook, no judgement calls — and change nothing except how tall candle 1 was:

Four identically-named patterns, a 7.50× spread in what they pay, and 46.15 percentage points of break-even win rate between the top row and the bottom. If you screened for “bullish engulfing” you would have taken all four without noticing they were different trades. Nothing in the name, the shape or the description distinguishes them.

The four-second test

Rearranging R:R = (D − x) ÷ (d + x) for a target multiple k gives a rule you can run before you have talked yourself into anything:

Apply it to our structure and something awkward falls out. The deepest close that still leaves a 2R trade is x = 1,667, which is p = 0.833. That is below the engulfing threshold. On this chart, with this stop and this target, a bullish engulfing cannot be a 2R trade at all — only a piercing line can. The 2R setup here has no name in the textbook.

Generalised, a minimum engulfing needs D ≥ 3b + 2d = 7,000, while a minimum piercing line needs only 1.5b + 2d = 4,000. The engulfing demands 1.75× as much room from the market for the same read. And in field units, measured across every engulfing in the simulation, the median requirement for 2R was 2.31× the pattern's own height from the pattern to the next level, rising to 3.18× for 3R. That is the version to carry around: if the next real level is not at least a couple of pattern-heights away, the pattern is not the trade.

Is the tweezer the only one of the four that gets this right?

Yes, and the reason is structural rather than lucky. The tweezer measures something horizontal, and horizontal evidence does not get billed to your stop.

Compare what the two families ask candle 2 to prove. An engulfing asks it to travel a distance — and that distance, x, is charged to you twice, as shown above. A tweezer asks something completely different: our course material defines it as two adjacent candles making two lows, or two highs, at the same level, with candle 1 red and candle 2 green at a bottom, or the reverse at a top. Nothing about how far anything travelled. The claim is that the market reached a price twice and was turned away twice.

Notice what that does to the incentives. In an engulfing, “more convincing” means candle 2 closed further away, which widens your stop. In a tweezer, “more convincing” means the two extremes matched more exactly, which lets you place the stop tighter, just beyond a price two separate candles have independently refused. Signal strength and reward-to-risk move in the same direction. It is the only pattern in this lesson where they do.

Which is why the tweezer has the worse hole in its definition

“At the same level” is not a number, and nobody publishes one. Lesson 25 made this argument about the doji's body — that an undefined threshold is a definition-shaped hole a trader quietly fills with whatever lets them take the trade they wanted — and the tweezer has the identical hole in a more dangerous place, because here the threshold is the stop.

Loosen the phrase from a twelfth of a candle to a third of a candle and you find 2.84× as many tweezers. At the loosest setting one pair in five qualifies, which is not a pattern, it is a description of ordinary price. The right column is the one to look at, because it says what you are really claiming: at a 0.20% tolerance, “the market was turned away at the same price twice” means the two rejections happened somewhere inside a band a third of a candle wide.

The instruction that follows is short and unglamorous. Pick a tolerance, write it down, express it as a fraction of the average candle range on the timeframe you trade rather than as a fixed number of dollars, and never widen it on a day when nothing is setting up.

What does the volume rule actually filter?

Less than it looks like, and on the engulfing it is close to self-fulfilling — because the pattern's own geometry makes the volume test easier to pass exactly when the trade is worst.

Our course checklist is firm about it: candle 2's buy or sell volume must overwhelm and exceed the candle before it. As a principle that is right, and it is the principle Lesson 14 is built on — volume measures effort. The problem is the way the test is written. “Candle 2's volume > candle 1's volume” compares two totals, and totals scale with how far a candle travelled. An engulfing already requires candle 2 to travel further than candle 1's body. So part of the volume increase is not evidence of effort; it is the definition of the pattern showing up again in a second column.

Work it through with modelled figures. These volumes are illustrative, not measured from an exchange — they are here to show the shape of the effect:

The large engulfing wins the raw test by twice the margin and shows barely any real increase in effort — 1.03× against 1.37× — while paying roughly half as much or worse. A screen built on “engulfing plus high volume” is therefore mildly biased towards the trades that price badly. That is not a reason to drop the volume rule. It is a reason to divide.

One honest caveat, because the mechanism rests on an assumption. This argument needs volume to rise roughly with a candle's range, which is a well-known regularity and not a law. If your instrument shows the opposite — large ranges on thin volume, which happens in illiquid hours and thin pairs — then a tall candle 2 with high volume genuinely is unusual, and the raw test carries more information than this section allows. The way to know is to look at the volume bars on your own pair rather than to trust either version of the rule.

Can you avoid the penetration tax?

Mostly, yes — and this is the practical payoff of the whole lesson. The tax is not charged by the pattern. It is charged by the decision to enter at candle 2's close.

Read the algebra once more: R:R = (D − x) ÷ (d + x). The x in that expression is not really “how far candle 2 travelled”. It is how far above candle 1's close you got filled. Those are the same number only if you buy at candle 2's close. Get filled lower and x falls, whatever candle 2 did.

So take the worst row in the master table — the big engulfing, candle 2 closing at 65,000, x = 3,000, paying 0.86R — and instead of buying it, leave a limit order at the midpoint of candle 2, which is 63,500. If it fills, x = 1,500 and the trade pays 2.25R. Same pattern, same stop, same target, same read.

Notice what the midpoint entry actually is. It puts your fill at the price a shallower penetration would have closed at — and shallower penetrations are the top rows of the master table. Waiting for the pullback converts a large engulfing into a piercing line's price. That is why the effect is so large: you are not being clever, you are moving your entry to a different rung of the ladder we already priced.

The obvious objection is that the pullback may never come. So we measured it, in the same driftless random walk. After each bullish engulfing, we asked whether price traded back to candle 2's midpoint within the next N candles before it traded through the stop level:

Ten candles of patience filled 65.0% of the time. Three candles filled 54.5%; twenty filled 69.1%. And the bar you need to clear is tiny: at a 55% win rate, entering at the close earns +0.02R per pattern, while the midpoint entry earns +0.79R per fill, so the limit order only has to fill 2.72% of the time to match it. It filled 24 times more often than that.

Two footnotes that make the case stronger rather than weaker. First, in 18.9% of patterns price went through the stop level without ever offering the midpoint. Those are trades the close-entry took and lost, and the limit order simply never entered — the non-fills are not evenly split between winners and losers. Second, without assuming any win rate at all: at a 65.0% fill rate the midpoint entry carries 1.71× the reward exposure per pattern and 0.65× the risk exposure. Better on both axes, and the only assumption needed is that the conditional win rate is the same.

That last assumption is exactly where to be sceptical, and where the honest limits of this measurement sit. A driftless walk has no momentum by construction, so a genuine reversal that runs away without looking back is under-represented here. It is also possible that patterns which do pull back to their midpoint are systematically the weaker ones, which would lower the conditional win rate on the limit entry. The gap between 65.0% and 2.72% is wide enough that the direction of the answer is safe, but the exact number is not yours — log your own fills for a quarter and use those.

PRACTICE CORNER

Ten minutes, and it settles the argument this lesson makes better than any explanation. Open a 4-hour chart of any liquid pair on the exchange you already use, scroll back three months, and find five engulfing patterns — you will not have to search hard, since one turns up about every sixteen hours. For each one write down four numbers and nothing else: candle 1's close, candle 2's close (their difference is x), the lower of the two lows, and the next obvious level in the direction of the trade. Then compute (D − x) ÷ (d + x). Two things will happen. Most of them will price under 2R, and the ones that price best will be the least impressive-looking candles on the screen. Then, on the same five, mark candle 2's midpoint and check whether price came back to it — that is your own fill rate, and it is worth more than ours.

Affiliate disclosure: the links below are partner links. We may earn a commission at no cost to you. It does not change what this lesson says. Full disclosure.

Binance→ OKX→ Bybit→

How do you run the check, in order?

The order is the useful part, because the expensive step comes last and three of the five can end the check before you have looked at the pattern at all.

1. Is there a trend for this to reverse? Our course checklist puts it first for a reason: without a side in control there is nobody to overwhelm, and without room to the next level the arithmetic collapses — 1.60R becomes 0.40R in a 3,000-point box. Diagnose trend or range before anything else. This step is free.
2. Measure D before you name the pattern. Find the next real level in the direction of the trade and write down how far it is from candle 1's close. Do this first, while you have no stake in the answer — the distance to the target cannot be argued with, whereas your enthusiasm for a pattern grows the longer you stare at it.
3. Measure d, the stop. The lower of the two lows for a bullish pattern, the higher of the two highs for a bearish one, plus whatever buffer you normally use. Not the body. Not candle 2 alone.
4. Now measure x, and price it. R:R = (D − x) ÷ (d + x). Below your floor, it is a no, however textbook the candles look. If it is close, check the shortcut: a 2R trade needs the level roughly 2.3 pattern-heights away.
5. Then decide where to get filled, not whether to. If the close-entry prices badly and the midpoint-entry prices well, leave the limit at the midpoint and accept that a third of the time you get nothing. That is the cheapest 23 points of break-even win rate on this page.
6. Only last, look at the volume bar — and look at it per point of range, per Lesson 14. Raw volume on a tall candle 2 tells you mostly that candle 2 was tall.

Naming the pattern is not on that list. You can run the entire check without ever deciding whether the thing in front of you is a piercing line or an engulfing, which is a fair summary of how much the name is worth.

When is everything above wrong?

Several of the numbers here come from a simulation and one of the arguments rests on an assumption, so it is worth being exact about the boundaries.

The simulated rates are accident rates, not win rates. A driftless random walk has no trends, no participants and no memory, so 12.521% is how often the shape appears when nothing is being decided. Nobody is right 12.521% of the time. What the number buys you is a floor: a shape this common by accident cannot be self-evidently meaningful.

Real markets trend, and trending markets are not the floor. In a genuine downtrend, red candles are larger and more frequent, which changes how often a green candle can cover one. The direction of every conclusion in this lesson survives that — the algebra in section 4 has no statistics in it at all — but the specific decimals do not travel to your instrument.

Volatility was fixed at 0.4% per candle. Change that and the tweezer tolerance table in particular moves, since it compares a fixed percentage band against a candle's typical range. On a thin altcoin or in the ten minutes after a headline, the same 0.10% tolerance is a much smaller share of a candle.

The contiguity argument depends on continuous trading. Everything about impossible gaps applies to a 24/7 spot or perpetual pair on one exchange. It does not apply to instruments that close — CME bitcoin futures gap over a weekend, and equities gap overnight, which is exactly why the classical definitions were written with gaps in them in the first place. Thin pairs and periods of exchange downtime can also produce real gaps.

The 50% and 100% thresholds are conventions. They come from the classical candle literature and from our own course material, not from a measurement. Nothing about the algebra breaks if your school uses 60%; the point of the lesson is that whatever threshold you use is being applied to a quantity that does not determine your trade.

And the fill-rate case assumes the same win rate on both entries. If patterns that pull back to their midpoint are systematically weaker ones, some of the 23-point advantage is repaid in a lower hit rate. The margin is large, but it is not infinite, and your own records are the only thing that closes that question.

What are the most common mistakes here?

Treating the pattern as the reason for the trade. Our course notes name reading a signal off one to three candles as an account-emptying habit, and this is a two-candle signal. It is a timing device inside a read you already had, which is why the checklist starts with the trend and not with the candles.

Waiting for candle 2 to cover candle 1's wicks. The course material says explicitly that candle 2 need not cover the wicks. Requiring it means a larger x, which means a wider stop and a nearer target — you are paying for strictness that the definition never asked for.

Believing the bigger engulfing is the better trade. It is the stronger claim and, on the same read, the weaker trade. 3.33R against 0.44R across the master table, and the name does not distinguish them.

Screening for the pattern on a 15-minute chart. Twenty-four a day counting both directions, in a market with nobody in it. There is no reading skill that recovers signal from a bucket that size.

Calling a gapless setup a dark cloud cover and then trading it as though the trap element were present. The trap in the classical pattern is the gap. Without it you have a bearish half-penetration, which is a real setup with a real price — just not the one the description is describing.

Using a fixed dollar tolerance for tweezers. “Within twenty dollars” means something quite different on a quiet 4-hour candle and a violent one. Express the tolerance as a share of the average candle range on your timeframe.

Entering at the close out of habit. This is the expensive one, and it is invisible because the trade still looks like the trade. The midpoint limit costs you a third of your entries and buys 23 percentage points of break-even win rate.

What else do people ask about two-candle reversals?

Is a bullish engulfing a reliable buy signal on its own?

No, and the arithmetic says why rather than just asserting it. On a continuously traded pair candle 2 opens where candle 1 closed, so the definition reduces to one comparison: did candle 2 close above candle 1's open? In our 400,000-candle random walk that happens on 12.521% of adjacent pairs — about 24 a day counting both directions on a 15-minute chart. Trends do not turn 24 times a day, so the shape cannot be the signal. What makes it tradeable is the two things our course material insists on: the market has to already be trending, and the trade has to price out. In a 3,000-point range instead of a trend, the same minimum engulfing pays 0.40R and needs a 71.43% win rate to break even.

Why does the piercing line I see in crypto never have a gap?

Because it structurally cannot. The classical piercing line needs candle 2 to open below candle 1's low, and the classical dark cloud cover needs it to open above candle 1's high. Both need a market that stops trading. A 24/7 pair never stops, so each candle opens at — or within a tick of — the previous close: in 399,974 adjacent pairs, the number that opened outside the previous candle's range was zero. This is why our own course checklist drops the gap and keeps only the 50%-penetration rule; that is the version of the pattern that can exist on a continuous chart. Gaps do still appear on instruments that close, such as CME bitcoin futures over a weekend.

Which is stronger, a piercing line or an engulfing?

The engulfing makes the stronger claim and is usually the worse trade, and both are true at once because they measure against different things. The name grades x against candle 1's body. Your reward-to-risk measures x against the room left to your target, and candle 1's body does not appear in it at all. On the structure in this lesson a minimum piercing line pays 3.33R and a minimum engulfing 1.60R off the identical read, stop and target. For a 2R trade the engulfing needs the next level 7,000 points away against 4,000 — 1.75× as much room from the market.

Where exactly does the stop go on a two-candle pattern?

Beyond the extreme wick of whichever candle reaches further: the top of the higher high for a bearish pattern, the bottom of the lower low for a bullish one. That is the right level, because it is the price that has to be wrong for your read to be wrong. What nobody mentions is how the stop divides. Risk is d + x, and in our data the median engulfing had x worth 70.2% of the total — so about seven-tenths of your stop is not protecting a level, it is paying for how far candle 2 happened to run.

Should I enter at the close of candle 2 or wait for a pullback?

Wait, on the numbers, as long as you accept that waiting sometimes means no trade. Entering at the close of a large engulfing here pays 0.86R and needs 53.85%; a limit at candle 2's midpoint pays 2.25R and needs 30.77%. At a 55% win rate the limit only has to fill 2.72% of the time to match the close entry, and in our data it filled 65.0% of the time within ten candles. The caveat that matters: a random walk has no momentum, so a real reversal that runs away without looking back fills less often than that. The margin is wide enough that the direction survives — but measure your own fill rate rather than trusting ours.

Educational content only — not financial advice, and not a trade recommendation. Pattern definitions, the entry/stop/target placement and the trending precondition are taken from our own slide course, Part VI. The two marked-up charts are from that course. Every figure on this page is either reproducible arithmetic from the worked structure (candle 1 open 64,000 close 62,000, pattern low 61,500, target 68,000) or comes from our own simulation: a driftless random walk of 400,000 candles, 60 sub-steps each, volatility 0.4% per candle, seed 20260903. The volume figures in section 7 are modelled to illustrate a mechanism, not measured from an exchange. Published 3 September 2026.