Four attempts, and the one that held

AttemptHighCloseSessionDaily buy share
1st, Aug 24$80,000$78,993+1.62%49.8%
2nd, Aug 25$81,273$78,5390.57% lower51.0%
3rd, Aug 28$81,479$77,8463.00% lower47.1%
4th, Sep 3$82,300$81,270+5.08%54.7%

The first three all made a higher high than the one before and all gave it back inside the session. The fourth made a higher high again and kept most of it, closing $2,424 above the third attempt's close.

The move ran through the European and US afternoon. The 14:00 UTC hour gained 2.17% on $200.4 million of spot at a 62.1% buy share, and the 15:00 hour added 0.99% on $246.9 million at 57.0%. Reporting attributes the trigger to Federal Reserve Governor Christopher Waller saying he is inclined to hold rates at the September 15 and 16 meeting if incoming inflation data confirms price pressures are easing.

The metric earned its keep, with a caveat

This site has spent two weeks documenting what the taker buy share cannot see. Intellectual honesty runs both ways, so here is a case where it did exactly the job you would want from it.

Set the four readings against the distribution we published on August 27. The daily median is 49.0% and the 90th percentile is 52.4%. The three failed attempts read 49.8%, 51.0% and 47.1%, all sitting close to a typical session. Thursday read 54.7%, comfortably above the 90th percentile and among the strongest daily readings the market produces.

Four events that looked similar on a price chart, and the metric separated the one that worked from the three that did not. That is a genuine result and we are not going to bury it under caveats.

We will, however, state the obvious caveat once. Four observations is four observations. This site has spent a month refusing to build rules out of handfuls, most recently on Sunday when a rare sequence appeared seven times in 999 sessions and we declined to say what followed the prior six. The same discipline applies to a result that flatters us. One clean discrimination is an anecdote, not a signal.

And open interest says something else

If the story ended there it would be a clean one. It does not, because our second test contradicts the first.

SessionPriceOpen interestReading
Aug 31+1.16%1.26% lowerCovering
Sep 11.45% lower1.56% higherPositions opening
Sep 20.13% lower0.29% higherRoughly flat
Sep 3+5.08%0.60% lowerCovering
Sep 4, partial0.48% lower4.71% higherPositions opening

On the day the breakout happened, open interest fell. Contracts were being closed while the price rose 5.08%, which is the same signature we measured on August 19 and named a short squeeze at the time. By this test the fourth attempt succeeded on the same fuel as the first leg of the August rally.

Then look at the last row. Since the break, open interest has jumped 4.71% on the daily reading and 4.57% over the last 30 hours, while the price has drifted slightly lower. Positions are being opened after the move rather than during it.

What to do when two of your own tools disagree

The honest answer is that they are measuring different things and both readings are correct.

The buy share counts who crossed the spread. On Thursday that was overwhelmingly buyers, and 54.7% across a whole session is a genuinely strong figure. Open interest counts whether contracts were created or destroyed. On Thursday they were destroyed on net, which means a large share of that aggressive buying came from people closing short positions rather than opening long ones.

Both facts describe the same session. Buyers were in control of the tape, and a meaningful part of what made them buy was that they had to. A short being closed is a real buy order that really lifts the offer, and it is also a buyer who will not be buying again tomorrow.

So the disagreement is not a malfunction. It is two questions with two answers: who was urgent, and whether exposure grew. A reader who only watches one of them gets a confident half of the picture, which is worse than knowing the picture has two halves.

What neither test can tell you is what happens next, and we are not going to invent an answer by combining them.

The level itself moved

One detail that most coverage of this zone has missed, including our own until we recomputed it today. The 50-week moving average that has been described as the ceiling is not standing still.

Measured on50-week averageChange
Aug 26, completed weeks to Aug 23$81,823
Today, completed weeks to Aug 30$81,071$752 lower

The average has fallen $752 in nine days, because higher weeks from last year keep rolling out of the fifty week window. The ceiling is descending toward the price, not waiting for it.

Against the current figure: Thursday's close of $81,270 was 0.25% above the average, and Bitcoin at $80,871 now sits 0.25% below it. A weekly close above $81,071 would be the first since May 4. This week's candle currently stands at $80,962 with two sessions left. We are stating where things are, not what they will do.

Positioning, and a setup we have now flagged three times

DateLong/short accountsContextWhat followed
Aug 20, before0.9988Below parity into a levelA $2.7 billion short squeeze
Aug 24, at the high0.9459Further below parityNothing. It unwound quietly
Today0.7979Far below bothNot our call to make

Today's reading of 0.7979 means roughly 56% of accounts are positioned short, immediately after a breakout to a four month high. It is by a wide margin the most short skewed figure we have recorded in this run. Open interest has risen 4.57% over 30 hours alongside it.

The third column of that table is the reason we are not going anywhere with this. We have flagged a sub parity reading twice. The first time a squeeze followed within a day. The second time nothing followed at all. Two instances, opposite outcomes, which is our own evidence that this configuration does not forecast.

Funding sits between 0.0038% and 0.0089% across the last six settlements, comfortably off the 0.0100% ceiling it was pinned to a week ago. Longs are not paying an unusual premium to hold.

Where things stand

AssetPrice24h24h range
Bitcoin$80,871+4.08%$77,478 to $82,300
Ether$2,505.77+4.30%$2,390.00 to $2,529.84
XRP$1.4440+5.56%$1.3589 to $1.4835
Solana$103.51+2.78%$99.93 to $105.91
BNB$723.59+4.12%$694.80 to $729.90
Dogecoin$0.08677+4.59%$0.08253 to $0.08999
Tokenised gold$4,468+0.63%$4,424 to $4,512

Everything is up and gold is up least, at 0.63% against four to five percent for the majors, which points at something crypto specific rather than a broad discount rate move. Total market capitalisation is $2.730 trillion, up 0.93%, with Bitcoin dominance at 59.33%. The price sits 16.2% above the 200-day moving average of $69,604.

On flows, August finished with about $3.52 billion into US spot Bitcoin ETFs, their best month of 2026, and September 2 added $101.15 million. Strategy also ended a two month pause, buying 4,603 Bitcoin for roughly $370 million between August 24 and 30, taking its holdings to 845,050 coins.

Fear and Greed reads 74, Greed, the highest of this run. The ETH to BTC ratio is 0.03098, up 0.23%, and it has now dropped below 0.031 for the first time in the ten days we have been reporting it. Ten observations, still not a trend.

What we are not saying

We are not saying the breakout holds. Bitcoin has already slipped back below the 50-week average it closed above on Thursday, and the week has two sessions left.

We are not saying a 54.7% reading predicts anything. It described Thursday accurately and separated it from three failures, on a sample of four.

We are not saying the short positioning will be squeezed. We have flagged that setup twice before and it resolved in opposite directions.

We are not saying which of our two tests is right about Thursday. They measure different things and both are correct as measurements.

What we are saying is that the fourth attempt cleared a zone that had rejected three, that our headline metric distinguished it while our positioning metric did not, and that the level everyone is watching has quietly fallen $752 in nine days. Every figure came from a public endpoint and every one is reproducible.

Earlier coverage, merged into this page

We covered this story day by day as it happened. Those daily pieces are now consolidated here, and their addresses redirect to this page. The finding from each, as published on the day:

  • Aug 24: Bitcoin Touched Exactly $80,000, and the Five Minutes After Read 19.9%. The 15:30 UTC candle high was $80,000.00 to the cent and closed below it at a 60.7% buy share. The next five minutes took $48.9 million at 19.9%. Across six sessions and a 26% move the daily buy share never left the 48.9% to 53.9% band, and its correlation with daily returns over 399 sessions is 0.505.
  • Aug 26: Bitcoin Gave Back the $80,000 Break, and 16 of 999 Highs Stopped on the Exact Dollar. Tuesday took Bitcoin through $80,000 to $81,273 and the session still closed lower. Measuring 999 completed daily candles, 16 highs landed on an exact $1,000 multiple against a uniform expectation near 0.001%, with the same effect in Ether, Solana and XRP. Right now the largest resting ask on the book sits at exactly $79,000.00.
  • Aug 28: Bitcoin Reached $81,479 and Gave Back 4.3%, and Open Interest Rose on the Way Down. The third attempt above $80,000 made the highest print of the run and the deepest give back of the week. Open interest rose 3.00% while price fell 2.57%, which is contracts opening rather than longs being liquidated. The claim that August was all forced short covering does not hold for the two biggest sessions.