Estimated liquidation clusters across price and time, rebuilt from Binance USDT-M open-interest flows. Bright horizontal bands mark price zones dense with leveraged positions; when price sweeps through a band, those positions are forced to close and the band is cleared.
Levels are model-based estimates built from Binance USDT-M perpetual open-interest changes distributed across assumed leverage brackets (10x / 25x / 50x / 100x), not actual liquidation orders. Brighter cells mark price zones where more leveraged positions would be forced to close.
Every leveraged perpetual position carries a liquidation price: the level where the exchange force-closes it because margin has run out. Individually those levels are private, but in aggregate they cluster at predictable distances from where positions were opened. A liquidation heatmap paints those clusters over time: the horizontal axis is time, the vertical axis is price, and the color intensity of each cell estimates how much position notional would be liquidated if price reached that level.
Clusters act like magnets and fuel. Market makers and momentum traders know that pushing price into a dense band triggers a cascade of forced orders, which is why price often accelerates into bright zones and stalls after consuming them. Bands above the current price are short liquidations (forced buying when hit); bands below are long liquidations (forced selling).
Exchanges do not publish other traders' liquidation prices, so the map is a model, not a record. Kodexius records Binance USDT-M open interest every 5 minutes. When the number of open contracts rises during a candle, each new contract is treated as one fresh long and one fresh short, opened near the candle's average price (open, high, low and close averaged). Their notional is distributed across assumed leverage brackets (10x, 25x, 50x, 100x), and each bracket's liquidation price is worked out with the exchange's isolated-margin formula and an assumed 0.4% maintenance margin rate. That notional is added to the price bins at those levels. When price later trades through a bin, 95% of its accumulated liquidity is cleared, mirroring the liquidations that would have fired there.
Only positions opened inside the selected window are modeled, so the left edge of a window starts empty and fills as open interest builds. The newest candles can show no new liquidity yet, because open-interest readings are published a few minutes after the fact.
Because raw estimates are dominated by a few giant clusters, the color scale is clipped at a fraction of the window's maximum value, the liquidity threshold (0.85 by default, adjustable above the chart). Lowering it makes smaller clusters brighter; raising it reserves the hottest colors for the very largest ones. The slider only rescales the coloring: the underlying data never changes, and the brightest color means "at or above the clip value", not a precise dollar amount. Treat the map as a relative density gauge, not an order book.
Source data is public Binance USDT-M perpetual market data: klines for price, and open-interest history collected every 5 minutes (Binance serves only the latest 30 days of it, so Kodexius stores each reading as it is published). The heatmap recomputes on demand and is cached for about a minute. Nothing on this page is investment advice.