The same fifteen minutes, measured on two books

Every figure in this section was computed from Binance candle data at the time of writing, from closed five minute candles, using the taker buy quote volume divided by the total quote volume. That is the same calculation the live dashboard runs. The only change is that we ran it twice, once on the spot pair and once on the perpetual contract.

08:05 to 08:20 UTCSpot BTCUSDTPerpetual BTCUSDTRatio
Value traded$140.8 million$1,973.4 million14.0x
Taker buy share63.0%64.2%almost identical
Net aggressive buying$36.5 million$561.6 million15.4x
Share of the combined net6.1%93.9%

The middle row is the one that matters for anyone using this site. The buy share is a ratio, and as a ratio it did its job perfectly. It told you that roughly two thirds of the value crossing the spread was buying, and that was true on both venues to within about a point. Someone reading only the spot number would have correctly concluded that buyers were in control of that quarter hour.

They would also have had no way to know that the buying they could see was about a sixteenth of the buying that was actually happening.

Why the shorts were standing there

Nothing was published at 08:05 UTC. There was no release, no statement, no data print. That is a European morning hour and a Tokyo evening one. So the honest question is what fired, and the answer is a price level and the positions parked around it.

Yesterday's story on this site was the Treasury doubling its long bond buyback operations, which took Bitcoin to a close of $69,335. We reported the price. What we did not do the maths on until today is where that close sat relative to the 200-day moving average, which stood at $69,035.

It was above it. That was the first daily close above the 200-day average since November 2, 2025, ending a run of 289 consecutive closes below it. We computed that from 1,000 daily Binance closes, rolling the average forward one day at a time, rather than comparing today's price to today's average, which is the common way to get this wrong.

A level that has held for two hundred and eighty nine sessions collects orders above it. Some of those orders belong to people who are short and have decided in advance where they will give up.

The positioning data says they were there. At 08:00 UTC the Binance global long/short account ratio on BTCUSDT read 0.9988, which is to say slightly more accounts were short than long, the only reading below parity in the surrounding sessions. By 09:00 it read 1.0777. In one hour the crowd went from marginally short to comfortably long, and it did not get there by changing its mind.

Open interest is how you tell covering from conviction

This is the part that separates a squeeze from a rally, and it is measurable rather than rhetorical.

When new buyers open fresh long positions, open interest rises, because a new contract creates a position on both sides of the trade. When existing positions are closed, open interest falls. So the direction of open interest while a price is rising tells you whether you are watching new money arrive or old positions surrender.

Hour, UTCBTC price movePerp open interestChangeLong/short accounts
07:00$69,852 to $69,816110,090 BTC+0.74%1.0101
08:00$69,816 to $71,592110,370 BTC+0.25%0.9988
09:00$71,592 to $71,790108,284 BTC1.89% lower1.0777
10:00$71,790 to $71,996108,177 BTC0.10% lower1.0338

Price rose 2.54% in the 08:00 hour and open interest fell 1.89% across it. Roughly 2,086 Bitcoin worth of open exposure disappeared while the price was climbing. That is not new conviction being expressed. That is old positions being taken off the board, and since the flow was aggressively buying and the accounts flipped from short to long, the positions coming off were shorts.

The industry figures put a size on it. Short liquidations across crypto reached about $2.7 billion in 24 hours, which CoinGlass records as the largest forced closure of bearish positions since its data starts in 2021. Total liquidations approached $3 billion across roughly 172,108 traders, with shorts about 92% of the damage against roughly $257 million of longs, a ratio above ten to one.

The third blind spot, and the running record

Six days ago this site published a piece arguing that our buy share could not see an $853 million ETF week, because ETF creations settle through authorised participants and never cross the Binance spread. Today produced a second, different kind of blindness, and it is worth setting them beside each other rather than treating each as a one off.

Kind of flowWhere it executesWhat our spot buy share showsCase
ETF creations and redemptionsAuthorised participants, off the public bookNothing. Direction and size both invisibleAug 14, Aug 16, Aug 17
Futures liquidationsPerpetual order bookDirection right, size understated about fifteen timesToday
Ordinary spot demandThe book we measureDirection and size both correctThe normal case

These fail differently and that difference is useful. An ETF week gives you a neutral spot reading while price climbs, which is confusing but at least it flags itself: the number looks wrong for the move. A liquidation cascade gives you a reading that looks completely reasonable. Sixty three percent buying on a sharp move up is exactly what you would expect to see. Nothing about the number tells you that you are looking at a sixteenth of the story.

The practical rule we are adopting from today: when Bitcoin moves more than about two percent in an hour and the spot buy share is merely elevated rather than extreme, check whether open interest fell. If it did, the size you are looking at is not the size that moved the price.

What today was not

Yesterday every risk asset on the board went up together, gold and silver included, and we said that pattern pointed at the discount rate rather than at anything to do with blockchains. That test is worth running again today, because today it fails.

Tokenised gold on Binance is up 0.55% over 24 hours. Yesterday it was up about 3.5%. Bitcoin is up 5.56%, Ether 11.22%, XRP 18.13%. When crypto moves five to eighteen percent and gold moves half a percent, the thing that changed is not the price of money.

The monetary news, in fact, went the other way. The FOMC minutes from the July 28 and 29 meeting were released at 18:00 UTC yesterday, which was after we published. They showed the hawkish side was larger than the vote suggested. Three regional presidents dissented in favour of a quarter point hike, which was already known, but the minutes and subsequent reporting indicated that Kansas City's Jeffrey Schmid and St Louis's Alberto Musalem, neither voting in July, would also have backed an increase. That is five officials leaning towards higher rates, against a 9 to 3 hold.

Crypto rose anyway, and it rose while gold sat still. Both of those facts point the same way: this was a positioning event and a policy headline, not a macro repricing.

One correction of emphasis on yesterday's piece, while we are here. We wrote that the Treasury had fixed the long end. The official Treasury close for August 19 confirms the 30-year at 5.19%, down from 5.28%, so the move was real. But Thursday's official close is not published at the time of writing, and market reporting through the session described long yields as rising again. One day does not fix a fiscal problem, and we should have said "relieved" rather than "fixed".

CLARITY, for the fourth time, is a calendar story

The headline everyone attached to this move was President Trump convening crypto executives and regulators at the White House on August 19 and urging the Senate to pass the CLARITY Act, calling for "a fair version" of the bill.

On August 2 this site argued that crypto legislation moves on the calendar rather than on its content. On August 3 we wrote that CLARITY needs seven Democrats and that Bitcoin was rising on the attempt, not the outcome. On August 17 we extended the same logic from Congress to the SEC when it cancelled its own vote. Today is the fourth instance and nothing about it breaks the pattern.

What changed yesterday was that the President asked. What did not change: the bill is still deadlocked in the Senate over ethics provisions that would restrict public officials from profiting personally from crypto, Republicans still resist those provisions as targeted, and the Senate is in recess until September. No vote has been scheduled. No Democrat has moved.

So the position we have held for three weeks stands unchanged: the market keeps paying for the attempt. It has now done so four times without a law existing. We are not going to pretend the fourth time means something the first three did not.

The flows our spot tape never records

Separately from anything forced, the unforced demand also arrived, and it arrived on the settlement rails we have spent a week explaining that we cannot see.

August 19 net flowAmountNote
US spot Bitcoin ETFs$517.19 millionLargest single day since May 4
BlackRock IBIT$284.7 millionAbout 55% of the total
ARK 21Shares ARKB$77.7 million
Fidelity FBTC$62.4 millionEight of twelve funds positive
US spot Ether ETFs$189.15 millionLargest since October 28, 2025
BlackRock ETHA$122.12 millionAbout 65% of the Ether total

Two details are worth keeping. First, eight of the twelve Bitcoin funds took money, which is a different picture from one fund absorbing a single large ticket. Second, the Ether figure is the strongest since October 2025 and lifts the month to about $534 million, the best of the year for those funds.

And one caution we have earned the hard way. Yesterday's article had to report that Monday's flow was first published as $137.3 million and finally settled at $297.5 million once BlackRock's data was included. Early ETF prints are partial. Treat any same day figure as provisional until the consolidated number lands.

Where things stand

AssetPrice24hNote
Bitcoin$72,370+5.56%High $72,830, the highest since June 1
Ether$2,320.97+11.22%Above any close since May 11
XRP$1.2588+18.13%Above any close since June 1
Solana$87.26+6.75%
BNB$650.96+5.52%
Dogecoin$0.07986+10.20%
Tokenised gold$4,512+0.55%The one that did not join

Total crypto market capitalisation is $2.474 trillion, up 3.05%, with Bitcoin dominance at 58.56% and Ether at 11.30%. The Fear and Greed index reads 62, Greed, against 46 yesterday, 41 on Tuesday and 31 on Monday. That is a thirty one point swing in four sessions, which says more about how fast sentiment surveys follow price than about anything else.

The ETH to BTC ratio is 0.03210, up 5.0% on the day. We reported it above 0.03 yesterday and we are reporting it again today. Two observations are not a trend and we will not describe them as one.

On the funding side, one number argues against reading euphoria into this. The BTCUSDT perpetual funding rate settled at 0.0094% at 16:00 UTC, essentially the venue's baseline. After a 5.56% day and a record short squeeze, longs are not paying an unusual premium to stay long.

Volume figures for August 20 in this article cover 00:00 to just past 16:00 UTC and are therefore partial. On that partial basis, spot BTCUSDT has traded $2.06 billion at a 51.2% buy share while the perpetual has traded $16.52 billion at 53.1%. Note what those whole day numbers do to the story: averaged across seventeen hours, both venues look like a coin flip. The entire character of the day lives in fifteen minutes.

What we are not saying

We are not saying the rally continues. A squeeze is buying that has to happen whatever the buyer believes, and it ends when the positions are gone. Roughly $2.7 billion of shorts have already been closed, which means that particular fuel has been spent rather than stored.

We are not setting a level. This site retired directional thresholds on August 3 after watching two of our own if then conditions fire in opposite directions within three days. The buy share describes the window it measures. It does not forecast the next one, and today it did not even measure the whole of the window it appeared to.

We are not calling the 200-day average reclaimed as a state of affairs. One close above a level that held for 289 sessions is one close. It is a fact worth knowing and it is not a trend.

What we are saying is narrower and, we think, more useful. On a day when our own headline metric would have led a careful reader to a reasonable but badly undersized conclusion, the correction came from a second measurement on a second venue and from open interest. That is a method, not a prediction, and it is available to anyone with the same public API.