Liquidation heatmap.

How this map is built — and what it can't tell you

When you trade with leverage, the exchange closes your position automatically if price moves far enough against you. The price where that happens depends on two things: where you opened, and how much leverage you used. Because most traders pick from a handful of common settings — 10x, 25x, 50x, 100x — their liquidation prices bunch up at predictable distances from wherever the crowd was opening positions. Those bunches are what this map shows.

No exchange publishes where open positions sit; that is private account data. So this is a model, and every liquidation heatmap you will find elsewhere is one too. Ours works like this: each hour, the increase in open interest tells us roughly how much new money entered at that hour's price. Perpetuals are symmetric — every long is matched by a short — so that new size splits evenly between the two sides, then spreads across the common leverage tiers. Each resulting liquidation price lands in a price bucket, and the buckets are what you see.

One step matters more than the rest: clearing. A level only stays on the map if price has not already been there. If price later fell through a long liquidation level, those longs are already gone and the level is deleted. That is what makes this forward-looking — a picture of leverage that is still alive and still exposed, rather than a history of what already blew up.

Read it as pressure, not prophecy. A large cluster is a place where reaching the price triggers forced selling (below) or forced buying (above), which pushes price further the same way — which is why clusters often behave like magnets. But the leverage mix is assumed, not known; traders close and hedge positions manually; and one large account can matter more than a thousand small ones. The map tells you where a move would find fuel. It does not tell you that the move is coming.

Methodology — the exact model behind this map

This map is rebuilt from Binance USDⓈ-M perpetual data: hourly candles plus open-interest history. Open interest arrives already denominated in USD notional, and every figure on this page is USD notional. No exchange publishes where open positions would actually be liquidated — that is private per-account state — so this is a model, and the whole model is below.

New leveraged exposure is attributed from the rise in open interest, not from volume. For each hourly candle, newly opened notional is taken as max(0, OI value now − OI value previous hour). Volume would double-count, because it counts closes as well as opens; when open interest falls, positions were closed, so nothing is added. Perpetuals are symmetric — every long has a matching short — so that newly opened notional splits 50/50 between the two sides.

Each side is then spread across four leverage tiers, weighted by an assumed share of open notional: 10× at 30%, 25× at 30%, 50× at 25% and 100× at 15%. A long opened at price P with leverage L liquidates near P × (1 − 1/L + MMR); a short near P × (1 + 1/L − MMR). MMR is the maintenance-margin rate, 0.4% — Binance's tier-1 rate for BTC and ETH. Those levels are accumulated into 56 price buckets spanning the observed range.

Then comes the step that makes this forward-looking rather than a museum of old positions: clearing. A long level at price X, created in one candle, is already gone if any later candle traded at or below X — it was liquidated when price got there. Shorts are cleared the same way against later highs. Triggered levels are dropped, so what survives on the map is leverage that is still live and still exposed. Clusters below spot are longs that would cascade on a fall; clusters above spot are shorts that would cascade on a rise.

What this cannot tell you. The tier weights are an assumption, not a measurement — no venue publishes the leverage distribution of open positions, and shifting those weights moves the clusters. The maintenance-margin rate used is tier 1, so unusually large positions sitting in higher margin tiers liquidate slightly earlier than modelled. Coverage is Binance USDⓈ-M only, so leverage stacked on other venues is invisible here. And attributing new notional to the hour in which open interest rose is an approximation, because opens and closes net out inside that hour. Every liquidation heatmap you can find is an estimate of the same unpublished state; the difference here is that you can check the arithmetic.

Frequently Asked Questions