What a liquidation heatmap actually shows
A liquidation heatmap looks like a measurement and is not one. Nobody — not this site, not an exchange, not the largest data vendor in the market — can see where other traders would be liquidated. Every heatmap in circulation, including the one on this site, is a model of an unobservable quantity. That is not a disclaimer bolted onto the picture; it is the first fact about it.
Why the underlying quantity cannot be observed
A position's liquidation price is a function of three things: its entry price, its leverage, and the collateral behind it. All three are private to the trader and the venue. An exchange knows its own book and publishes none of it position by position; a venue that settles on-chain makes each position readable in principle, but only by address, which means reconstructing the market would require indexing every account on every venue and then knowing which accounts matter.
What is public is aggregate open interest — the total value of contracts open on a contract — and the venue's own margin rules. A heatmap is what you get when you take those two public facts and assume the missing one.
How the picture is built
| Step | What happens | What is assumed |
|---|---|---|
| 1 | Take the contract's open interest — a real, published figure. | Nothing. This is measured. |
| 2 | Split it into slices, each representing positions opened at a given leverage. | The leverage mix: what share of the book sits at 2×, 5×, 10× and so on. Assumed, and the single largest source of error. |
| 3 | Place each slice at the price where a position of that leverage, opened around the prices that were trading when it was opened, would be closed. | That positions were opened across the recent price range rather than all at once, and that the venue's published maintenance margin applies. |
| 4 | Sum the slices into price bands and shade by density. | That older positions have partly been closed already, so the model decays them rather than carrying every position forward forever. |
Step 2 is where every heatmap differs from every other one, and almost nobody publishes their answer to it. Ours is below.
The leverage assumptions this site uses
Three profiles, selectable on the map itself. Each is a distribution of notional across leverage levels; the shares sum to one, and levels above a contract's own maximum leverage are redistributed rather than silently dropped. These are assumptions, published so that a reader can disagree with a specific number rather than with the picture in general.
| Profile | 2× | 5× | 10× | 20× | 25× | 40× |
|---|---|---|---|---|---|---|
| Aggressive | 5% | 11% | 18% | 20% | 24% | 22% |
| Balanced | 9% | 19% | 27% | 22% | 15% | 8% |
| Conservative | 23% | 32% | 25% | 12% | 6% | 2% |
Aggressive: Most size at high leverage. Closest to what a retail-dominated perp market is usually assumed to look like. Balanced: Weight spread across the range, tilted slightly to the middle. A deliberately unopinionated default. Conservative: Most size at low leverage, as a market dominated by funded desks rather than retail would look.
How to read one without being misled
| A common reading | What the data supports |
|---|---|
| "There is $400M of liquidity sitting at that level" | There is no liquidity in the picture at all. A heatmap contains no orders — it estimates where forced closing would occur, which consumes liquidity rather than providing it. |
| "Price is going to that cluster" | The map has no directional content. It says where a move, if it happens, would meet forced selling or buying that could extend it — a conditional, not a target. |
| "The bright band is where liquidations happened" | The map is forward-looking and hypothetical. Liquidations that have already happened are events, and no venue publishes a complete feed of them — why we publish no totals. |
| Comparing the absolute dollar values between two sites | Meaningless without both leverage assumptions. Two models over the same open interest can differ several-fold purely from step 2, and only one of the two usually publishes it. |
What it is genuinely good for
One thing, and it is worth having: identifying price regions where forced closing would be self-reinforcing. If a large share of the position base would be closed within a narrow band, a move into that band mechanically produces more of the same move. That is a structural observation about how the book is built, and it holds under a wide range of leverage assumptions even though the dollar figure attached to it does not.
Read as regions rather than lines, and as relative density rather than absolute dollars, a heatmap is a useful map of fragility. Read as a measurement, it is a confident-looking picture of something nobody can see.
Liquidation totals are a worse number than the map
Most sites that publish a heatmap also publish a "total liquidated" figure, and that number has a harder problem: the exchange feeds it is summed from are throttled, so the total is systematically understated, and most understated during the cascades that make it newsworthy. This site publishes no such total, and the mechanism is set out in full on why we don't publish liquidation totals. A modelled map that says it is a model is a more honest object than a measured total that is not a total.