Liquidation Heatmaps Explained — What They Show and What They Don't

What is a liquidation heatmap?

A liquidation heatmap is a chart overlay that estimates the prices at which leveraged positions would be force-closed by an exchange, colouring each price band by how much notional size is projected to sit there. Bright bands mark prices where a large amount of modelled exposure would be liquidated at once; dark areas mark prices where little would. It is the output of a model built from open interest, price history and assumed leverage — not a record of orders that exist on any exchange.

The critical word is estimated. Nothing on a liquidation heatmap has been observed. No exchange publishes a list of open positions with their liquidation prices, so every heatmap on the market infers where leverage was probably added and at what ratio. The map is a hypothesis about positioning, rendered with the visual confidence of measured data. That mismatch between how it looks and what it is causes most of the bad reads.

Used correctly, it answers one question well: where would forced flow be manufactured if price went there? It does not answer where price is going.

How is a liquidation heatmap estimated?

The construction is mechanical once you accept the assumptions. A liquidation price is not mysterious — it is the price at which a position's margin falls below the exchange's maintenance requirement, and it follows directly from the price the position was opened at, its leverage and its margin mode. The modelling problem is that only one of those three is inferable from public data.

A typical pipeline runs like this:

1. Track open interest changes against price. When open interest rises over a bar, new positions were opened somewhere in that bar's range. The bar's price range becomes the assumed opening range for that new notional.
2. Assign a direction. Whether the new exposure is long or short is inferred, usually from the sign of aggressive flow or price displacement in the same window. This step is already a guess.
3. Spread the notional across leverage bands. The same block of new open interest is duplicated at 5x, 10x, 25x, 50x, 100x and so on, each copy projected to the liquidation price that leverage implies. This is why heatmaps show several parallel bands trailing away from the same origin — they are the same position modelled at different leverage.
4. Age and decay the map. Levels are removed once price trades through them, since positions there are assumed liquidated, and often decayed over time to account for positions closed voluntarily.
5. Render as intensity. Overlapping projections at similar prices sum into bright clusters.

What no liquidation heatmap can see

The single most useful thing to understand about the tool is that a liquidation heatmap contains no positions. It is derived from aggregate open interest — one number per venue per contract, total notional outstanding — which says nothing about who holds it, at what price, at what leverage, or against what collateral. Every band on the map is reconstructed from that aggregate by assumption.

What is observed versus what is inferred:

Three consequences follow, none of them defects of a particular provider. Brightness partly encodes an assumption: the bands closest to price come from high-leverage assumptions and the distant ones from low-leverage, so a cluster's visual weight reflects a belief about trader behaviour as much as a measurement of size. The map is path-dependent: it is rebuilt from the price history that actually happened, so two tools with slightly different open interest sampling draw visibly different maps of the same market — compare the same symbol across two vendors before trusting either. Executed liquidations are a different, partial dataset: Binance's public stream pushes at most one order per second per symbol, so a cascade of hundreds of closes appears as a handful of prints and any cross-venue total built from public streams is a floor, not a count.

None of this makes the heatmap worthless. It makes it a positioning prior: a statement about where leverage plausibly sits, carrying error bars that no colour scale displays.

How do you read a liquidation heatmap?

Read it as a map of conditional fuel rather than a set of targets, and take the reads in order of reliability: the asymmetry between the two sides first, then cluster density relative to distance from price, then what the tape does if price actually arrives. The phrasing that keeps you honest is always conditional — if price reaches this band, the forced closes there are market orders in the direction of the move, which can extend it beyond what discretionary flow alone would produce. Nothing on the map tells you whether price will get there.

In more detail, in that order of reliability:

• Asymmetry between sides. A market with dense modelled long liquidations below and sparse short liquidations above is describing a positioning imbalance. That is a statement about how the book is leaning, and it is the most robust thing the heatmap offers.
• Cluster density versus distance. A cluster far from price requires a large move to matter and will likely be rebuilt or decayed before it is reached. A dense cluster close to price is the more actionable observation, because the move required to trigger it is one that ordinary flow can produce.
• Behaviour on arrival. This is where the heatmap stops being useful alone. When price enters a cluster you switch to the tape: does CVD spike with matching displacement — flow being taken — or does aggression arrive and price stall, which is absorption and means someone was waiting there with size?
• What happened after a sweep. A cluster that gets run and immediately reclaimed is a different event from one that gets run and holds as resistance. The heatmap cannot distinguish them; the reaction can.

Do liquidation clusters attract price?

No. There is no mechanism by which resting leverage pulls price toward it — a cluster is not an order, and nothing in the matching engine routes flow in its direction. What you should not do, then, is treat a bright band as a price objective.

Clusters get reached because ordinary buying or selling took price there. The leverage only changes what happens once it arrives: forced closes convert into market orders pushing the same way, which is why a move through a dense band often travels further than the flow that started it. That is a conditional amplifier, not a magnet.

Liquidation heatmap versus order book liquidity heatmap

These two overlays look similar and are routinely confused, including in tool marketing. They are close to opposites.

The practical consequence: liquidity heatmaps show you where a move might stop, liquidation heatmaps show you where a move might accelerate. When a dense book level and a dense liquidation cluster sit at the same price, you have a genuine confluence of two independent readings — and also the setup where the outcome is most binary, because either the resting size absorbs the forced flow or it does not.

Are liquidation heatmaps reliable?

They are directionally useful and precisely wrong. The open interest and price inputs are real, but the leverage attached to each position is assumed, cross-margin and hedged accounts have liquidation prices that no public data can recover, and traders add margin or close out before the level is ever touched. Treat a bright band as a zone where forced flow is plausible, not as a level that will be hit or held. Reliability degrades with two things in particular: distance from current price, and the size of the participant.

Three failure modes account for most of the error.

Leverage is assumed, not known. The single largest source. A position's liquidation price is a direct function of its leverage, and no public feed reveals it. Every band on the map is "this size, if it used this ratio."

Cross-margin and hedges are unmodellable. A cross-margin position's liquidation price depends on the whole account's equity — including unrelated positions and any collateral the trader deposits mid-trade. Hedged books may have no meaningful liquidation price at all. Both are common among the largest participants, which means the model is weakest exactly where the size is.

Traders intervene. Margin gets added, stops get moved, positions get closed before the liquidation price is touched. Every one of those makes a modelled cluster evaporate silently, with nothing in the data to mark it.

Which liquidation heatmap tool should you use?

Everything above argues for the mature providers rather than rolling your own. CoinGlass covers the widest set of centralised venues and is the sensible free starting point; Hyblock Capital goes further on leverage-band modelling and cohort segmentation for paying users. If your question is "where is aggregate leverage clustered across exchanges", either is a better answer than anything you would build alone, and we would point you there rather than at us. For a side-by-side of what each vendor models differently, see CoinGlass alternatives and how to interpret them.

How the Confluence Engine treats liquidations

We do not run a liquidation heatmap as a standalone signal, because a projection built on assumed leverage cannot carry a thesis by itself. In the Confluence Engine, forced flow enters as a layer among more than forty: realised liquidation prints from the Hyperliquid tape as observed events, alongside open interest changes, funding, spot-versus-perp CVD disagreement and the footprint detail of what actually traded at the level.

That combination is what makes a liquidation read meaningful. A cluster reached on rising open interest with aggression converting into displacement is a different auction from the same cluster reached on falling open interest with flow being absorbed — same price, opposite implication. The heatmap alone cannot tell them apart; order flow can.

And whatever the layers say, the output is always a scenario with its invalidation stated in advance: while price holds above X, the bias is Y. See how the engine works, or watch the show free on a delay.

Liquidation heatmaps: frequently asked questions

What is a liquidation heatmap? A chart overlay that estimates where leveraged positions would hit maintenance margin and be force-closed, coloured by how much notional size is projected to sit at each price. Bright bands mean a lot of modelled exposure is clustered there. It is built from open interest and assumed leverage — not a list of orders sitting on an exchange.

How do you read a liquidation heatmap? Start with the asymmetry between the two sides: dense modelled longs below and sparse shorts above describes a positioning imbalance, and that is the most robust read available. Then weigh cluster density against distance from price. When price actually enters a cluster, stop reading the heatmap and read the tape — order flow tells you whether the forced closes are being absorbed or are extending the move.

Are liquidation heatmaps reliable? Directionally useful, precisely wrong. Open interest and price are observed; leverage is assumed, margin mode is unknowable, and traders add collateral or close early. Use the bands as zones of plausible forced flow with error bars, not as levels.

Do liquidation clusters pull price toward them? No. A cluster is not an order and nothing routes flow toward it. Clusters get reached by ordinary buying or selling; the leverage only changes what happens after arrival, when forced closes add same-direction market orders and can extend the move.

Educational analysis, not financial advice. @TheConfluenceShow