Bitcoin has risen 22.9% since Monday’s open. Over the same stretch, the number of bitcoin-denominated futures contracts open on Binance — the largest venue by a wide margin — has fallen 5.7%. Funding has not once, in the entire move, printed above the venue’s baseline rate. This is the strangest 20%-plus rally we have had to write about, and a week of extra data has made it stranger, not less so.

Yesterday we published this as a finding with a caveat: the coin series and the dollar series disagreed, and we said which one we quoted and why. Today the disagreement is wider and the direction is unchanged. Here is the full week.

Reading one: open interest, in coins

Own pull, Binance fapi openInterestHist (daily) and openInterest (live), 22 August 2026.

Read the two numeric columns against each other. In coin terms open interest went 111,988 → 105,618, a fall of 5.7%. In dollar terms it went $7.057bn → $8.308bn, a rise of 17.7%. Same market. Same instrument. Opposite stories.

The dollar series is not lying, it is just doing arithmetic you did not ask for. Notional value is contracts multiplied by price. When price rises 23% and contract count falls 6%, notional rises about 18% and a headline writes itself: “open interest surges to $8.3 billion as bitcoin rallies.” That sentence is true and it means the opposite of what the reader will take from it. The number of bets on the table went down. Every one of them just got marked up.

This is now rule one of our short-squeeze field guide and it is the single most useful thing on this page: read open interest in coins, or do not read it at all.

Reading two: funding never left the floor

A perpetual futures funding rate is what longs pay shorts (or the reverse) to keep the contract tethered to spot. When leveraged longs crowd in, funding rises — that is the mechanism by which you can see a rally being bought on borrowed money. Binance’s baseline for BTCUSDT is 0.0100% per eight-hour period. Here is every settlement since the move began.

Own pull, Binance fapi fundingRate, BTCUSDT. Eight consecutive settlements. Maximum: the baseline itself.

Not one settlement above baseline. Bitcoin rose $9,003 across those eight windows — from Wednesday’s $69,334.79 close to Friday’s $78,338.03 — and the derivatives market never once charged a premium to be long. In the leverage-driven rallies people remember — the ones that end badly — this column reads 0.05%, 0.10%, occasionally 0.30%. Here it reads like a market where nobody had to pay to participate because there was no queue.

Reading three, and the new one: the crowd sold it, the big accounts bought it

Binance publishes two positioning ratios that are usually quoted interchangeably and should not be. The global long/short account ratio counts accounts — one retail wallet, one vote. The top trader long/short position ratio weights by position size among the largest accounts on the venue. This week they went in opposite directions, and that is the observation we did not have yesterday.

Own pull, Binance fapi globalLongShortAccountRatio and topLongShortPositionRatio, BTCUSDT, daily period.

The retail account base entered the week 2.22 accounts long for every one short and finished it at 0.96 — net short, on Friday, into the highest bitcoin price since May. It has since crept back to 1.04. Over the same four days the largest accounts on the venue went 1.49 → 1.43 → 1.50 → 1.72 → 1.96, rising on each of the last three readings and finishing at the highest print of the series.

Two populations, one tape, exactly inverted. The crowd that was positioned long into a flat market got shaken out or took profit on the way up; the size accounts added into strength. We are not going to tell you which group is right — the top-trader ratio is a notoriously good contrarian signal at extremes and a notoriously good trend signal in between, and anyone claiming to know which regime this is on a Saturday morning is guessing. But it is a clean, checkable divergence, and it is the reason we are no longer comfortable calling this a purely retail-driven squeeze.

So what actually moved the price?

Two readings survive the data, and we are going to publish both rather than pick, as we did yesterday.

Reading A — there was nothing to flush, so it went up. Aggregate bitcoin futures open interest had been grinding lower for a year, from a peak around $45 billion in July 2025 to about $20.4 billion by 30 June 2026, in what analysts cited in market commentary described as an orderly deleveraging rather than a crash. A market that has already expelled its leverage has almost no supply of forced sellers. Add a real catalyst — the Treasury buyback, the White House crypto roundtable on 19 August, $1.9 billion of ETF demand — and price has to travel a long way to find sellers. Under this reading the move is healthier than a leveraged one and has further to go.

Reading B — the fuel is spent. Bloomberg’s headline figure was roughly $2.7 billion of short liquidations in a single 24-hour window (Wednesday into Thursday), inside a ~$2.99bn day-total across 172,108 traders, of which shorts were about 92% against roughly $250–257 million of longs. Across the two-day stretch, short liquidations alone are reported above $4 billion. Two caveats we owe you: these are reported CoinGlass figures, not our own pull, and CoinGlass is known to understate totals because Binance has throttled its liquidation feed since 2021. Either way it is a one-off. Every short that was going to be squeezed has been squeezed. The bid that carried the last $6,000 was mechanical, not discretionary, and mechanical bids do not repeat. Under this reading Monday opens without its engine.

The honest position is that the same three data series support both. What they do not support is the third story — that leveraged longs are piling in and the market is fragile. Coin open interest is down 5.7%, funding is at baseline, and the retail crowd is flat. Whatever risk sits under this rally, it is not the risk of a crowded long.

The counterweight we are obliged to print

Bernstein’s read of the week is that this was a liquidity event: the Treasury’s buyback expansion functioned as a monetary easing signal and every risk asset with a duration story attached repriced. That is broadly consistent with what our own data shows.

The other side of the argument, and the one worth holding on to, comes from MEXC’s Shawn Young, who has argued that crypto is giving the Treasury’s intervention far more credit than it deserves, and named two invalidation levels: the 10-year above 4.7% and the 30-year above 5.3%. That is not a distant hypothetical. The 10-year closed Friday at 4.736% — already through the first level — and the 30-year at 5.276%, within three basis points of the second. If the liquidity story were the whole story, those two yields should have fallen after the buyback. They rose. The rally and the bond market are currently disagreeing about what Wednesday meant, and only one of them can be right.

Marker board

Y2 — PASS, with a deflator we owe you. Both legs cleared, and neither was close: bitcoin finished 24.3% above its bar and the S&P 1.4% below its own. A marker the tape clears by a quarter is not a test of anything. We wrote it on 20 August with bitcoin at $69,335 and it was already 10% in the money the moment it was published. That is a badly-sized marker and the PASS is worth very little. We are recording it as such.

Z2 — the one that matters, and it is failing in our favour. Z2 was written deliberately so that a PASS would destroy our own central claim: if open interest rebuilt to its pre-rally level, the “de-levered rally” thesis would be wrong and we would have to say so. It has not rebuilt. It has fallen further, from 109,390 at Friday’s reading to 105,618 this morning — the lowest coin-denominated print of the entire series. Six days remain on the marker. We would rather report this than a marker we could not lose.

Correction: the difficulty retarget lands Sunday, not Saturday

In yesterday’s edition we wrote that U2 settles at approximately 23:56 UTC on Saturday 22 August. That was wrong. The current estimate from mempool.space is 00:15 UTC on Sunday 23 August — about 19 minutes later, and on the other side of midnight, which also puts it on the other side of the weekly close. We regret the error and we are flagging it rather than quietly changing the number.

The seven-snapshot projection series, taken at roughly the same hour each morning, reads: −1.63%, −0.28%, +0.49%, +0.37%, −0.94%, −1.20%, and this morning between −1.22% and −1.32%. Two sign flips in six days. Four calls to the same endpoint inside twenty minutes this morning returned −1.3159%, −1.3159%, −1.2417% and −1.2246%, which tells you the noise floor is roughly nine basis points of adjustment and that quoting three decimal places would be false precision. 104 blocks to go. We will grade on the settled print and write the post-mortem on how badly a mid-epoch projection predicts a settlement, because that is what the series was for.

The counter-signal we are still carrying

Network hashrate fell hard into the rally: daily averages of 1,025.3 EH/s on Monday, 1,000.2, 941.1, then 782.1 EH/s on Thursday — a 23.7% three-day drop into a 21% three-day price gain. It has since recovered to 812.7 on Friday and 863.6 on Saturday, with an instantaneous reading of 913.8 EH/s. The dominant explanation in current reporting is the AI and high-performance-computing pivot among listed miners, with CoinShares projecting that AI/HPC could account for roughly 70% of listed-miner revenue by the end of 2026; Texas grid curtailment is real but secondary. Either way, the retarget arriving Sunday is the network’s own accounting of it.

What would change our mind

Three things, in order of how much they would cost us. First, Z2 passing — coin open interest back above 111,988 — would mean leverage rebuilt and the thesis was a description of a moment rather than a structure. Second, funding printing meaningfully above 0.0100% for two or more consecutive settlements would mean longs are now paying to be there. Third, the top-trader ratio continuing above 2.0 while the account ratio stays below 1.0 for another week would mean the divergence is a positioning regime rather than a squeeze artefact, and we would have to stop calling this a squeeze at all.

Monday brings the first institutional flow print since the high, Bessent’s sanctions detail, and the grades on Z1 and U2. Wednesday brings July PCE. Friday brings Warsh at Jackson Hole and three markers settling at once.