What the data actually said

Two releases landed together at 8:30am Eastern, which is 12:30 UTC, and they pulled in opposite directions on the surface while pointing the same way underneath.

Core PCE, the gauge the Fed actually targets, eased to 3.3% year over year in June from 3.4% in May, matching what economists expected. The monthly figure was 0.1%, cooler than the 0.2% consensus. Headline PCE printed 3.7%, also in line. So inflation moved in the right direction, modestly, from a level that had been the highest in about three years.

Then the growth number. Advance second quarter GDP came in at an annual rate of 1.5%, against 2.1% expected and 2.1% actually delivered in the first quarter. That is a six tenths miss and a clear deceleration. The Bureau of Economic Analysis attributes the increase to consumer spending, investment and exports, partly offset by a decline in government spending.

Put those together and you get the combination that argues for patience: inflation drifting down, growth slowing faster than expected. It is not a strong economy that needs cooling. It is a slowing economy with a lingering price problem, which is a much harder thing to hike into.

The bond market did not wait for the data

This is the part that gives yesterday's decision its real cost. The Fed held, and the long end of the curve revolted.

The 30-year Treasury yield rose 10 basis points to 5.21%, reported as its highest in 19 years. The 10-year added 7 basis points to 4.67%. And the 2-year fell 4 basis points. Read that shape carefully, because it is the whole message: the short end came down because no hike arrived, and the long end went up because traders decided inflation will therefore run hotter for longer. That is a steepening driven by credibility, not by the rate path. The bond market was not pricing a different Fed. It was pricing a Fed that will be later.

Equities took it badly. The Dow fell 1,153 points, about 2.1%, its worst session since April 2025. The S&P 500 lost 1.5% and the Nasdaq 1.7%. Ellen Zentner of Morgan Stanley Wealth Management summarised the positioning shift plainly, saying market pricing for a hike had been pushed forward and that September remains a live meeting.

Chair Kevin Warsh gave the hawks no comfort in tone even as he outvoted them. He described the three dissents with the phrase "I asked for a good family fight and I got one", insisted there is no soft target and only a 2% target, and noted that years of above target inflation cannot be cured in nine weeks. He also brushed off a soft June consumer price reading as not much of a consideration for him. Fifteen hours later the June PCE confirmed exactly the cooling he had declined to lean on.

Our own markers, checked honestly

Yesterday's article named three things to watch. All three resolved inside 18 hours, and the bearish one triggered first. That order matters, so here it is straight.

The warning fired. We wrote that the 18:05 candle printing 48.0% was the tell, and that if the following hours kept coming in below 50% while price held, it meant pre positioned longs distributing into their own good news and the level would eventually give. The 19:00 UTC hour fell 1.41% on $93.1 million. Looking inside it, there was no single violent candle: buy shares oscillated between 42% and 60% and the price simply bled lower for twelve straight five minute bars. That is what distribution looks like when nobody panics.

The downside level broke. We named $63,562 as the marker whose loss would mean the post decision bid had failed entirely. The low reached $63,267 in the 21:00 hour. It broke. What did not happen is the next clause of that sentence: we said losing it would bring $62,500 back into the conversation, and the market stopped $767 above that.

Then the upside confirmation arrived. We wrote that clearing $64,744, the morning high and the honest resistance rather than the statement spike, on a buy share above 60% would confirm the hold was being bought rather than sold into. At 11:50 UTC today a five minute candle cleared it with a 74.1% buy share on $9.4 million. Bitcoin has since printed $65,176 and trades near $64,892, up about 1.43% on the day.

So the framework got the sequence right and the magnitude wrong in one direction. The break happened as described. The extension to $62,500 did not. Both are worth recording, because a method that only publishes its hits is not a method.

The pattern that repeated: buying before the release

Here is the finding worth keeping. Two scheduled macro releases, 18 hours apart, and the same signature in both.

Window (UTC)EventVolumeTaker buy share
Jul 29, 17:00 to 17:55Hour before the Fed statement$31.1M65.2%
Jul 29, 18:00 to 18:10After the Fed statement$39.1M55.0%
Jul 30, 11:30 to 12:25Hour before PCE and GDP$59.0M64.6%
Jul 30, 12:30 to 13:25After PCE and GDP$32.1M57.8%

Read the buy share column downward. 65.2%, then 55.0%. Then 64.6%, then 57.8%. Two independent events, and in both cases the aggressive buying was materially stronger in the hour before the number than in the hour after it. The pre-release windows are not noise either: the 12:10 candle today ran an 84.7% buy share on $11.4 million, and 11:50 ran 74.1%.

One difference is instructive. The Fed statement produced a $32.4 million five minute candle, 12.5 times the prior average. The PCE and GDP release produced a $4.7 million candle, which is entirely ordinary, at a 73.0% buy share. In other words the second release generated no volume event at all. The market had already decided, and the number simply confirmed it without argument.

Two observations do not make a law. But the practical implication is the same either way: on a scheduled release, the hour that carries information is the one before the clock, not the one after.

Where the real absorption happened

The single largest hour of the whole stretch was not around either release. It was 09:00 UTC today, which carried $165.1 million, and its buy share was only 48.3%. Sellers were marginally the aggressive side on the biggest volume of the day, and the hour still closed 0.40% higher. Inside it, the two largest five minute candles were 09:05 at $30.1 million on a 49.8% buy share and 09:45 at $20.5 million on 43.8%.

That is textbook absorption. Aggressive sellers crossing the spread in size, into resting bids that do not move the price down. It is the least glamorous reading on this page and probably the most important one, because it is where the supply that had built up since the statement actually got cleared. The reclaim at 11:50 only became possible after that hour ate the offers.

The tape has not been one sided since. The 14:00 hour ran a 38.9% buy share on $69.1 million and gave back 0.52%, a genuine seller push in the middle of an up day. The 23 hour aggregate sits at 52.7% on $1.293 billion. Note the volume: yesterday the comparable stretch was $816 million, so participation has risen by roughly 58% while the buy share stayed almost exactly where it was, at 52.8% then and 52.7% now. More people trading, same balance of aggression.

The weak link is still the ETF desks

None of the above changes the institutional picture, and it deserves an honest airing because it is the strongest argument against everything the flow is saying.

Spot Bitcoin ETFs have taken in about $205 million net in July, described as the lowest monthly total on record, with two trading days left. That number needs its context to be read correctly, and the context cuts both ways. It follows $2.43 billion of outflows in May and $4.52 billion in June. So July is the first positive month after two badly negative ones, which is an improvement, and it is simultaneously the feeblest positive month there has ever been. Both statements are true and neither is comfortable.

Meanwhile Ether ETFs pulled in $342.85 million in July, more than Bitcoin's total, and that is the more interesting line. XRP funds added $13.61 million on track for a fourth straight month of inflows, and Solana funds $13.82 million. Within crypto, allocation is rotating away from Bitcoin at the margin even as Bitcoin's own order flow holds a 52.7% buy share.

Two other honest notes. Gold outperformed on price again: tokenised gold is up roughly 2.3% over 24 hours against Bitcoin's 1.43%, though its buy share remains lower at 44.9% for Tether Gold and 50.8% for Paxos Gold. And sentiment still refuses to participate. The Fear and Greed index reads 28, one point lower than yesterday, with the last five readings at 26, 30, 29, 29 and 28. Total crypto market capitalisation is $2.296 trillion, up 1.48%, Bitcoin dominance 56.6%.

What today's data does to September

The next decision is September 15 and 16, and it brings back the Summary of Economic Projections and the dot plot after two meetings without one. Today's numbers changed the burden of proof inside that meeting.

Yesterday the three dissenters had a clean story: inflation has been above target for years, so tighten. This morning core PCE came down and growth missed by six tenths. Hiking into a 1.5% growth print is a much harder case to make than hiking into a 2.1% one, and each of them now has seven weeks of data ahead that could go either way. As Zentner put it, the hike looks postponed rather than cancelled, and September is live.

The bond market is the group that has to be convinced, not the committee. A 30-year at 5.21% is the market charging a premium for the Fed's patience. If the next two inflation prints keep drifting down, that premium unwinds and long yields fall, which historically is the friendliest macro backdrop a scarce asset can get. If inflation stalls near 3.3% while growth keeps slowing, the Fed is boxed in and the long end keeps rising, which is the version nobody in risk assets wants.

What to watch next

  • Whether $65,176 gets cleared. Today's high is the new marker. Taking it out on a buy share above 60% would extend the reclaim rather than cap it, and it is the first level above that has not yet been tested twice.
  • $64,744 as support now, not resistance. The level that took 18 hours to break should hold on the retest if the reclaim is real. Losing it on a sub 45% buy share would mean the whole recovery was a squeeze.
  • The last two ETF sessions of July. Two trading days remain to move a $205 million monthly figure. A strong finish reframes July as a turn; another outflow day makes the weakest positive month on record look like a pause on the way back down.
  • Whether the long end keeps rising. The 30-year at 5.21% is the single most important number for risk assets right now. It matters more than the next crypto headline.
  • The pre-release window on the next scheduled print. Two for two on the buying arriving early. A third instance turns an observation into something worth trading around.
  • Whether Ether keeps out-collecting Bitcoin. $342.85 million against $205 million in the same month is a rotation, and if August repeats it the dominance number will start to move.

How to read this as a trader

Three days ago the useful lesson was that flow shows damage before the news explains it. Yesterday it was that the buying can be finished before the announcement. Today the same thing happened again, at almost identical numbers, on a completely different release. That repetition is the actual finding, and it points at something structural rather than lucky: positioning ahead of a known clock is a decision made by people who cannot wait for the print, and it leaves a measurable footprint in the aggressive side of the tape.

What that means in practice is unglamorous. Around a scheduled release, the trade is not the reaction. By the time the candle prints, the people who were going to act have acted, and the volume you are seeing is often them handing inventory to whoever chased. Yesterday's 12.5 times volume spike at a falling buy share was that handoff in a single bar. Today the release did not even produce a spike, because the transfer had already happened in the previous hour at 64.6%.

The other lesson is about levels. Both of the markers we published yesterday were hit, but the bearish one hit first and the bullish one needed 18 hours. Anyone who treated either as a forecast rather than a threshold got whipsawed twice. A level is a place where you find out what the flow does, not a prediction about where price goes. Our live dashboard separates aggressive buying from aggressive selling minute by minute across the most liquid pairs, which is how the 09:00 absorption and the 11:50 reclaim were visible in real time. For the week's arc, see yesterday's decision and the first pre-release bid, the morning call that the sellers had stopped, Tuesday's broken bid, and Monday's reclaim of $65,000. If the buy and sell split is new to you, the volume guide covers the mechanics.