Liquidation Heatmap

THE MECHANISM IS GENUINE. A leveraged position is liquidated when its margin no longer supports it. Liquidation forces the position to be reduced or closed - for a losing long that creates selling flow, and for a losing short buying flow. Execution mechanics vary by venue: partial liquidation, immediate-or-cancel or limit orders, a liquidation engine, an insurance fund and auto-deleveraging are all used, so 'a forced market order' is a simplification. Those orders push price further the same way, which can trigger the next liquidation. That is a positive-feedback loop, and it is why crypto declines sometimes accelerate into vertical moves out of proportion to any news. It is the same reflexive dynamic as a margin call cascade in traditional markets, made faster by continuous trading and higher leverage.

WHAT DETERMINES A LIQUIDATION PRICE. Ignoring fees and venue-specific tiering, the adverse distance a position survives is roughly the INITIAL margin rate less the MAINTENANCE margin rate. For a 10x isolated position with a 0.5% maintenance rate that is about a 9.5% adverse move - not exactly 10%, and the shortfall is the maintenance buffer. At 20x it is about 4.5%. So a model placing hypothetical clusters 5-10% from spot is assuming concentrations around 10x-20x leverage: those distances are CONSEQUENCES OF THE ASSUMPTION, not observations of where positions sit.

AND HERE IS THE PROBLEM WITH ANY HEATMAP. Exchanges do NOT publish individual positions, their entry prices, or their leverage. They publish aggregate open interest and, some of them, liquidations after they occur. So a liquidation map cannot be observed - it must be MODELLED, by assuming a distribution of leverage across the open interest and projecting where those hypothetical positions would be liquidated. The clusters are outputs of that assumption, not data. Change the assumed leverage mix and the clusters move.

A STUDENT SHOULD BE ABLE TO SAY WHAT WOULD MAKE THIS OBSERVABLE: complete position, collateral and entry-price data, which most centralised venues do not publish - though transparent on-chain perpetual protocols expose considerably more of it - or realised liquidation prints, which exist but only AFTER the event and therefore cannot map what lies ahead.

A SECOND-ORDER PROBLEM. Widely published cluster levels become known to everyone, and the standard argument is that price is then drawn toward them as participants trade against the anticipated forced flow - a stop-hunt. If that is true, the map partly causes what it predicts, which makes it very hard to evaluate as a forecast.

WHAT THE INPUTS DO SUPPORT, stated more carefully than it usually is. Open interest is real: it counts contracts outstanding. But it counts CONTRACTS, not leverage and not capital committed - the same open interest can be held at 2x or at 50x - so its change is not a measure of how levered the system has become.

And the familiar rule that rising open interest with rising price 'means new longs' does not survive inspection: every newly created contract creates one long AND one short simultaneously, so open interest can never identify which side is new. Price-and-open-interest patterns are at best heuristic evidence about which side was the aggressor. Treat them as folklore with a grain of truth, not as identification.