Longs Paid Thursday, Shorts Paid Friday: Anatomy of a $1.3B Liquidation Week

Bitcoin traded between $76,651 and $79,837 in roughly 24 hours last week and liquidated close to $1.3 billion of positions in both directions doing it. Thursday belonged to the shorts, Friday belonged to the longs, and the price ended almost exactly where it started. Nothing about that is visible on a daily candle.

The trigger stack was macro. August producer prices came in 0.4 percent higher month over month on Thursday, bitcoin lost $77,000 and printed $76,651, and derivative liquidations peaked at $562 million with roughly $484 million of that on the long side. Nearly $74 million of longs went in the four hours immediately after the release.

Friday reversed it. August CPI landed in line with the 3.3 percent annual estimate, bitcoin spiked to $79,837, and total liquidations across the market topped $732 million with $424 million of wiped out shorts. Rate hike odds on prediction markets went from 63 percent before CPI to 82 percent after, which tells you the flush was mechanical rather than a change of thesis.

Why a crypto liquidation heatmap sees this before the candle does

A liquidation heatmap plots where leveraged positions get force closed, derived from position size, entry price and margin. Clusters form because traders enter at similar levels with similar leverage, so their stop out prices stack into bands.

Price does not drift through those bands. It accelerates into them, because each forced close is a market order that pushes price further into the next cluster. That is why an 0.4 percent inflation print moved bitcoin 3 percent in four hours while an in line print moved it 4 percent the other way.

The asymmetry between Thursday and Friday is the part worth internalising. Thursday drained long side fuel. That left the book unusually clean above spot, so when CPI removed the tail risk on Friday there was almost nothing between price and the short cluster sitting above. The second cascade was created by the first one.

Reading the tape while the cascade is running

Three orderflow reads separate a liquidation cascade from real directional selling, and they all work in real time rather than after the fact.

Cumulative volume delta measures net aggressive buying against net aggressive selling. During a pure liquidation flush CVD collapses at a rate completely out of proportion to the price move, then flattens the moment the forced sellers are done. Price falling while CVD stops falling is the classic exhaustion signature.

Order book imbalance shows whether resting liquidity is being pulled or stacked as price approaches a cluster. Market makers widen and step back when they expect forced flow, which is why depth thins out right before the fast move rather than during it.

VPIN, the flow toxicity measure, spikes when volume becomes overwhelmingly one sided. Elevated VPIN into a thinning book is the combination that precedes the violent leg, and it typically resolves within minutes of the last liquidation clearing.

Funding reset and what bitcoin flow says now

The cleanest evidence that last week was positioning and not repricing is where funding sits afterwards. Live Hyperliquid data on September 14 has bitcoin marked at $78,845, up 2.04 percent in 24 hours, with $2.87 billion of open interest and funding at baseline, 0.00125 percent per hour. Ether sits at $2,530 with $2.66 billion of open interest at the same baseline funding.

Baseline funding after a double sided flush means the crowd got knocked out of both trades and has not rebuilt a lean yet. Open interest holding up while funding normalises is the healthy version of that reset. The unhealthy version is open interest climbing while funding runs hot, which is what the book looked like going into Thursday.

Exact liquidation levels for tracked wallets, book fragility and VPIN across 530 plus pairs sit on buildix.trade/screener, which is free to start and updates on the same cadence the cascades develop on.

Two more inflation prints and a Fed meeting stand between now and the end of the quarter. The positioning that gets built in the next two weeks decides how violent those sessions are, and that positioning is measurable today rather than explainable afterwards.