BTC liquidation map

Hourly candles to 09:00 UTC, refreshed every 2 hours and drawn here in 2-hour steps. The dashed line is the live mark, $83,689, from the snapshot taken 2 min ago. Two different ages, so neither is allowed to stand for the other.

The assumptions, in full

Row 8 is the one most worth reading twice. Tying the axis to what actually traded is what gives the price path presence — the previous version spanned a third of the price to accommodate 2× levels half a market away, and flattened real movement into a squiggle. A minimum of ±8% stops a very quiet window from zooming into noise, and in a quiet window it is that minimum rather than the traded range that sets the axis. Either way the cost is that low-leverage levels fall off the picture. Here that is 2×, 5× and 10×.

Leverage mix used: Balanced

The exact weight vector behind the picture. The form below switches to any other profile in one submission.

The bars that did the clearing

The dark corridor in the picture is not empty space, and it was not carved evenly. These are the 6 bars of this window in which the model says the most notional was traded through — 72.04% of everything cleared in 14 days, in 6 bars out of 168. The largest was 21 Sep 2026 08:00 UTC, a 2-hour bar that ranged $81,767 to $85,500. The bar and its range are traded candles; what it cleared is the model's, and the two are separated in the columns below for that reason.

A bar clears a level by trading through it, so the column on the right is bounded by what the model had standing there in the first place — a violent bar through a thin part of the field clears less than a quiet one through a crowded part. That is the reading the picture gives at a glance and this table gives in figures.

Where the model puts the biggest clusters

Price levels still carrying modelled notional at the right-hand edge of the chart, largest first — $216.89M across the 10 listed. These are the numbers the bright bands stand for.

Where liquidation levels would sit if positions were opened the way the assumptions above describe — and, the part that matters, where price has already cleared them. The dark corridor the price path carves through the field is not empty space; it is the levels the market has already taken out. This is a model, and it is labelled as one inside the picture, not only underneath it. Every modelled figure here is scaled from Hyperliquid's open interest in BTC, which is one venue's book rather than the whole market's. What a liquidation heatmap actually shows →

How far price can move before each leverage liquidates — derived, not modelled

No assumptions here. Given the tier table and a mark these prices are arithmetic: the upper cap is where a short is closed, the lower cap where a long is closed.

Where BTC's margin tiers step down

Maintenance margin is set by the tier the notional falls in. Above $150.00M the corridor widens, because maintenance margin rises to 2.500%.

What is deliberately absent

No liquidation totals and no cascade tracker. Those need an event feed, and every available one is throttled — Binance publishes at most one liquidation per symbol per second, so a sum over it understates by an unknown factor, worst exactly when the number is most quoted. The full reasoning →

The 5 February sweep, on real candles → · Leverage survival — real candles, one stated rule → · BTC funding and margin tiers →

A page for every covered contract

55 contracts have a published margin table, and each one has its own page at its own address. The corridor column is not modelled: it is the distance price has to travel before a position at that contract's maximum leverage is closed by the maintenance margin in the published table — the widest of them is the safest place to be wrong, and the narrowest is where a normal hour is enough.

Maintenance margin at tier 1 sets every corridor above; the tiers themselves are on each contract's own page, together with its map and the assumptions behind it wherever its hourly history can be read and is long enough to draw one.