ETH: Estimated Liquidation Levels Open by Side (Calibrated)

Shown as its 2–31 Oct 2026 sample; the latest readings and full history are included with Plus.

How much in estimated long liquidations sits below the ETH price and how much in short liquidations above it, within 15% of the price, from the calibrated map, hour by hour.

Why use it. This is the stock of liquidation fuel on each side: the leveraged positions a fall or a rise of up to 15% would force out. It grows as leverage builds and shrinks when a move clears one side or traders close. How to read it. Each line is one side. A side climbing while the price goes sideways is leverage building up; a side collapsing during a move is its levels being liquidated. When one side dwarfs the other, a move toward it has more forced orders to feed on. The summary value is both sides together.

Calculation. For each period, the calibrated liquidation map at the period's end (the map the Futures Liquidation Heatmap (Estimated) draws): its estimated long liquidations and its short liquidations within 15% either side of the price, each side summed over the map's 0.5% bands and valued at the level prices. Long levels sit below the price and short levels above it, bar the odd level a venue's own price has not yet reached on the far side of the map's reference price (the mean of the venues' closes). Every estimated position opens as one long and one short level, so new positions add the same to both sides; the sides differ by what price moves have cleared. The map is a continuous estimate on a 0.05% price ladder rather than a list of positions, so the levels are measured by size, not counted. The map is a model, not an observation: no exchange publishes the leverage or entry price of open positions. Every five-minute rise in a contract's open interest counts as new positions opened at that bar's typical price (high + low + close) / 3, one long and one short per unit of open interest, spread over leverage tiers. A long's estimated liquidation price is entry x (1 - 1/leverage + 0.5%), a short's entry x (1 + 1/leverage - 0.5%) (0.5% maintenance margin). A fall in open interest closes the same share of every estimated position on that contract, and a level is removed once a bar's low (longs) or high (shorts) reaches it. Inverse and coin-margined contracts are counted by their USD face value. The calibrated map weights the tiers (2x, 3x, 5x, 10x, 25x, 50x and 100x) with a mix fitted to what we observe: how much Bybit, whose public feed reports every liquidation, actually liquidated when the price reached each tier's levels, and the liquidation prices of the large Hyperliquid accounts we track. Calibration version 2 (20 October 2026) fits the map's size with the mix: the USD Bybit liquidated over the USD the map estimated at the levels the price reached, one level for the two largest futures markets, bitcoin and ether, and one for every other coin, so the levels near the price are on the scale of the liquidations observed. It applies to the whole history. The fitted weights and levels, the fit's sample and their scores, with version 1 (the mix alone) kept as history, are published on the liquidation map accuracy page (/liquidation-map-accuracy/); a refit applies from its date on and earlier periods keep the version in force then. The accuracy charts score the calibrated map next to fixed weights (35% at 10x, 35% at 25x, 20% at 50x and 10% at 100x). Source: CryptoKeyhole's replay of our own collection of five-minute open interest and price bars of the ETH perpetuals on Binance and Bybit (USDT-margined, coin-margined and inverse) since 16 September 2026, and on Bitget and OKX, plus the USDC-margined perpetuals of Binance and Bybit, since 15 October 2026; before 16 September 2026 Bybit alone, replayed from the start of its public five-minute open-interest history (2020 for most contracts). Binance and Bybit open interest is each venue's own five-minute figure; for the contracts added later it is the median of our three minute samples ending at each five-minute mark. A venue's levels build up over the weeks after it joins, because positions opened before its data begins are not modelled; other venues' positions are not included, so the levels show the shape of leverage, not its total.