The compression resolved, upward and hard

Three sessions tell the whole story of the last four days.

SessionRangeAs % of closeChangeVolumeBuy share
Sat Aug 15$2680.42%0.07% higher$0.34B57.5%
Sun Aug 16$6741.07%0.29% lower$0.30B52.7%
Mon Aug 17$1,8592.88%2.59% higher$0.91B50.4%
Tue Aug 18, 4 hours old$5210.81%0.55% lower$0.11B48.6%

Saturday was the narrowest full session in three weeks at 0.42%. Monday was the widest since July 31, which ranged 4.68% on the day Bitcoin lost $62,500. Volume tripled from the weekend, from $0.30 billion on Sunday to $0.91 billion.

Bitcoin cleared $64,000 at 15:25 UTC, on a five minute candle carrying $18.9 million at a 60.5% buy share, and went on to a 24 hour high of $64,610. It trades near $64,177 now, up about 1.15% over 24 hours, with today's session four hours old and giving a little back.

Worth noting against what we published yesterday. The Monday previews had said Bitcoin needed $64,000, and at the time we wrote, its high was $480 short of it. It cleared the level in the US session a few hours later.

A coin flip produced the biggest move in three weeks

Here is the finding, and it is uncomfortable for anyone who wants a single number to explain a market.

Monday's taker buy share was 50.4%. Out of every thousand aggressive dollars that crossed the spread on Binance spot, roughly five hundred and four were buying and four hundred and ninety six were selling. That is as close to neutral as this metric gets. And the price rose 2.59%.

Look at where the volume actually was.

Hour (UTC)VolumeBuy sharePrice move
17:00$91.1M56.4%0.33% higher
15:00$86.8M49.9%0.47% higher
18:00$67.1M56.8%0.16% higher
05:00$49.0M45.1%0.02% lower
07:00$44.3M36.6%0.24% lower
16:00$43.1M42.1%0.13% lower

Every one of the three largest hours fell inside the US session, and the 15:00 hour, which produced the second largest gain of the day at 0.47%, did it on a buy share of 49.9%. That is sellers marginally ahead, with price up half a percent, on the second heaviest volume of the day.

This is the venue effect we set out on August 14, arriving in its clearest form yet. Our measurement covers one exchange's order book. When demand is being expressed through ETF creations, an OTC desk or any venue that does not lift offers on Binance, the split can sit at a coin flip while price runs. The Asia hours are the control: the 07:00 hour ran a 36.6% buy share on $44.3 million and price fell, which is exactly what a genuinely seller led hour looks like on this instrument. The US hours look nothing like that, and they are where the day was made.

The other series, published alongside as promised

Since August 14 we do not publish a buy share without the ETF flow beside it. Today that produces a split verdict and one honest gap.

The quarterly picture has turned. Yusuf Fakhro of ARP Digital reports more than 14,000 BTC of spot ETF inflows across the five days into August 7, the strongest stretch since May, and roughly 11,000 BTC of net inflows for the third quarter against about 110,000 BTC of outflows in the second half of the second quarter. His summary is that the institutional selling which defined that quarter has flipped to buying.

The weekly picture disagrees. The week of August 10 to 14 saw about $389.7 million leave, the largest weekly outflow in six weeks, and Bloomberg reported exactly that on Monday while the price was rising.

And the gap: Monday's own flow figure is not published yet at the time of writing. So we cannot tell you that ETF creations caused Monday's move. What we can tell you is that the move happened in US hours on a neutral exchange split, which is the signature of demand arriving off our venue, and that the most recent institutional data points in two directions depending on the window you choose. Anyone offering more certainty than that today is filling in a number they do not have.

Two figures in BTC terms are worth holding onto, because they are cleaner than dollar flows: 110,000 out in half a quarter, 11,000 back in so far. The turn is real and it is small.

The 30-year hit its highest since 2007 on the same day

The macro event that shared Monday with the breakout is one we have been following since the Fed decision.

The US 30-year Treasury yield reached 5.29%, reported as its highest level since 2007, and Bitcoin rose through $64,000 in the same session. Regular readers will recognise the instrument. On July 29 we wrote that the 30-year had risen 10 basis points to 5.21% after the Fed held on a 9 to 3 vote, that the 2-year had fallen 4 basis points at the same time, and that the steepening was a credibility trade rather than a rate path trade: the market pricing a Fed that would be later, not different.

What has changed is the co-movement, and the honest thing is to flag that we cannot yet say which version is the rule.

  • July 31. The long end sat at multi decade highs and Bitcoin fell 2.92%, losing $62,500, while tokenised gold fell with it. We called that a hedge unwind, and the evidence was that gold and Bitcoin sold together.
  • August 17. The long end made a fresh high at 5.29% and Bitcoin gained 2.59%, while tokenised gold was roughly flat.

Same variable, opposite responses, eighteen days apart. A rising long end can be read as an inflation problem that favours a scarce asset, or as a tightening of real financial conditions that punishes everything, and over the last three weeks this market has traded both readings. We have no basis for telling you which one holds from here, and after a fortnight of correcting our own extrapolations we are not going to invent one.

The correction: we gave the wrong date twice

On both August 16 and August 17 we listed the FOMC minutes as due on Wednesday August 20. That is not a date. August 20, 2026 is a Thursday.

The minutes for the July 28 and 29 meeting are scheduled for Wednesday August 19, which is tomorrow. Our weekday was right and our date was one day late. The error came from repeating a figure out of a preview that carried the same internal contradiction, and from not checking a weekday against a calendar, which takes one command.

Both articles have now been corrected, in English and in Arabic, and the wrong date appears nowhere on the site. We are stating it here rather than fixing it quietly because a reader who put our date in a diary would have looked for the minutes a day after they were published, and because a factual error about a scheduled event is exactly the kind a publication should be visibly accountable for.

Two series we refused to call trends, and both turned

This is the second consecutive article in which restraint adopted after an error has paid for itself, so it is worth keeping a scorecard.

The Ether over Bitcoin ratio. Yesterday we published that it stood at 0.029970 with six of the last seven daily closes higher, and wrote that we were reporting the sequence as a measured fact and specifically not calling it a trend, having extrapolated a five reading series in gold a fortnight ago and watched it reverse inside a day. It then reversed inside a day. Monday closed at 0.029653 and today reads 0.029511, partial. Ether is 0.48% lower over 24 hours while Bitcoin is 1.15% higher. The Ether leadership we described lasted one session.

The partial session label. On Sunday we published a range of $190 and marked it partial because the day had not closed. Sunday closed at $674, three and a half times larger. Without the label that would have been a fourth wrong characterisation in three weeks.

Neither of these is a clever call. Both are the absence of a bad one, which is less satisfying to write and more useful to have done. The gold error on August 3 cost us a whole section of a published article. The two rules that came out of it, label partial figures and refuse to name a trend from a handful of observations, have now each caught something inside a fortnight.

This was Bitcoin alone

Nothing else joined in, which is a change from yesterday morning when Ether led.

PairTaker buy share, 24h24h change
Bitcoin50.1%1.15% higher
Solana51.8%0.15% lower
Tether Gold54.5%0.07% lower
Paxos Gold50.9%0.18% lower
Ethereum48.2%0.48% lower
XRP45.6%1.26% lower
BNB42.2%0.34% lower

Bitcoin is the only pair higher on the day. XRP is down 1.26% and has slipped back below a dollar at $0.9907. Bitcoin dominance has risen to 56.6% from 56.2% yesterday, with Ether at 10.0%.

Sentiment finally moved properly. The Fear and Greed index reads 41, up ten points from 31 and the highest of this whole stretch after eight consecutive readings pinned between 27 and 34. It is still labelled Fear, and it is the closest to neutral we have recorded since we started tracking this run. Total crypto market capitalisation is $2.275 trillion, up 0.47%.

What is scheduled

Scheduled events and things we will measure, with no thresholds attached that claim to predict a direction.

  • FOMC minutes, Wednesday August 19, tomorrow. The written record of the 9 to 3 hold. What Hammack, Kashkari and Logan put on paper is the nearest read on whether September is still live for them after CPI at 3.4% headline and 2.5% core, PCE easing to 3.3%, GDP at 1.5% and a weak July jobs print.
  • Jackson Hole, August 21 and 22. Chair Warsh's first symposium in the job, and he has consistently preferred saying less about the policy path rather than more.
  • Monday's ETF flow figure, once it publishes. This is the single number that would settle whether the coin-flip breakout was fund demand arriving off our venue. We will report it either way, including if it undercuts the reading above.
  • The 30-year at 5.29%. The most important number for risk assets right now, and the one whose relationship to Bitcoin has inverted twice in three weeks.
  • September 15, still doubled. The reported CLARITY procedural vote and day one of the Fed meeting that brings back the dot plot.
  • Whether the SEC reschedules its cancelled crypto rulemaking vote. Still no new date announced since the August 14 cancellation.

How to read this as a trader

One lesson from Monday, and it is the same one three weeks of data keeps producing from different angles.

A neutral reading is not the same as nothing happening. If you had glanced at a 50.4% buy share on Monday you would have concluded the day was balanced and uneventful. It was the widest session in three weeks and it gained 2.59%. The number was accurate and the inference would have been wrong, because the metric answers a narrow question, which side paid the spread on one exchange, and Monday's answer to that question was genuinely close to even while the price was being set somewhere else.

The practical version is a habit. When the split is neutral and the price is not, do not conclude the market is quiet. Conclude that you are looking at the wrong venue, and go and find the flows you cannot see. On Monday those were in the US session, in the ETF wrapper, and in a 30-year yield making its highest print since 2007, and the exchange tape reported all three as a coin flip.

The corollary is the reason we now publish the two series together. A buy share alone would have described Monday as a non event. A flow figure alone would have described the week as distribution. Both are true, they cover different windows, and only the pair gets you close to what actually happened. Our live dashboard separates aggressive buying from aggressive selling minute by minute across the most liquid Binance pairs, which is one venue, precisely measured, published with its blind spot named. For the run-up to this, see yesterday's two regulatory stalls, the weekend where the venues cancelled out, the venue limitation in full, and the Fed hold that started the 30-year story. If the buy and sell split is new to you, the volume guide covers the mechanics.