The liquidation heatmap

Every leveraged position has a breaking point: a price where the exchange force-closes it. EdgeDepth maps that pressure across 660+ Binance USDT-M crypto and TradFi perpetuals, plus every Hyperliquid perpetual, then keeps three kinds of evidence distinct: a universal modeled field, covered real Hyperliquid open positions, and confirmed forced liquidations.

What you’re looking at

Four universal layers plus one Hyperliquid-specific layer. Each one is a toggle in the terminal’s LAYERS row.

Liq Heatmap, the field. The dense heatmap behind price. Every zone is colored by how much liquidation fuel is projected to sit there, on a ramp that runs from a near-black toe, through deep ember violet, up to a rare orange-and-yellow ignition. Bright means a thick cluster of leveraged positions that get force-closed if price arrives. It builds across the whole chart and moves with market history, not with your cursor, so a bright band stays pinned at its real price no matter how you zoom or scroll. Ordinary churn leaves only a faint trace; anomalous flow burns bright. The brightness is the signal, and the small notch on the legend marks where ignition begins.

The forward cascade. Look right of the last candle. Every zone still standing projects into the future behind a thin seam at the live edge, and its length is its strength: the heaviest magnets reach the far edge of the chart, while weak fuel is a short stub that fades out. One glance at the right side tells you what is still out there and how much it matters, a histogram drawn at the exact prices it lives at.

Liq Levels. The sharp horizontal lines drawn on top: modeled high-confidence clusters built from open interest, positioning and the prices where positions were added.

Liq Levels HL. Live price clusters built from covered real Hyperliquid open positions at the exchange-reported liquidation price. Actual leverage filters the layer, and the HL · COV badge shows how much venue open-interest notional the changing wallet census covers. Read exactly how the census works.

Liq Profile. The quiet silhouette down the right edge, the field collapsed into a single distribution. Most of it hugs the edge as a thin outline, and it bulges only where fuel has genuinely stood bright, so you see at a glance which prices hold the most standing fuel.

Fuel below price is long liquidations (leveraged longs getting stopped out); fuel above is short liquidations. A zone stays lit at the price where it formed, for hours, across every timeframe, until price reaches it from the side that triggers it. Long fuel below is spent when price falls to it; short fuel above is spent when price rises to it. The moment price trades through, that zone goes dark right there, and the carved channel traces price’s path through the chart. So fuel below survives a rally and fuel above survives a sell-off. When fresh leverage piles back in, the zone re-lights, brighter or dimmer depending on how much size just built. What is on the chart is always the fuel that is still there: never a graveyard of levels price already ate, and never a band smeared across the candles it already traded through.

How it works: two engines

1. The projection: works on everything, replays perfectly. We read the price and volume tape candle by candle. Each burst of activity is a cohort of leveraged positions opening near that price. We project it forward to where it breaks at every common leverage tier (25× to 100×) using the liquidation geometry, weight it by how anomalous the flow was (ordinary churn barely registers, while outsized volume marks where real leveraged size piled in), and accumulate it into the field over time. Standing fuel also ages: a zone nobody refreshes slowly dims toward the dark toe, while a zone that keeps attracting size stays bright. As price trades through a zone that fuel is consumed and the channel carves open; as fresh leverage rebuilds there, it re-lights. The grid itself is relative, so a coin that just did 8× renders as cleanly at the bottom of its range as at the top. Because it is computed purely from price action, the field renders on every symbol and every timeframe, including thin alts an open-interest feed cannot see, and it reproduces identically in replay, bar for bar, with nothing baked or stored.

2. The confirmation: real positioning, real positions, real liquidations. On liquid markets we layer live open interest, positioning, the trade footprint and forced-liquidation events on top. On Hyperliquid, a separate wallet census reads the exchange’s own liquidation price and actual leverage for each covered open position, filters unusable cross-margin values, and aggregates the remainder into Liq Levels HL. A dedicated Observed layer plots every confirmed forced liquidation as it prints, with bubbles sized by USD notional and tinted by side.

The field shows the universal pressure landscape. Modeled Levels summarize the strongest inferred clusters, Liq Levels HL adds direct but coverage-limited evidence, and Observed shows what actually fired. The sources are labeled and never silently pooled.

How to use it: the playbook

The heatmap is a map of pressure, not a prediction. Used right, it tells you where moves are likely to accelerate, where they are likely to stall, and where to place risk. Used wrong, it is a wall of pretty lines. The difference:

• Trade the cluster, not the line. A single level rarely matters; a dense stack of fuel is the magnet. Price tends to wick to the far edge of a cluster, where the cascade exhausts. That is the target, not the first line it touches.
• Read the imbalance as bias. Compare the long fuel below price to the short fuel above. The heavier side is the tank: when price breaks toward it, forced liquidations pour in the same direction and the move speeds up. A lopsided map is a directional tell; a balanced one (ordinary chop) means the map is context, not a signal, so size down.
• On a vertical pump, wait for the retrace. While price runs, the fresh fuel stacked just under it is untested and decorative. The edge comes when price falls back into that stack: it either cascades (fuel ignites) or the top holds as support. The re-test is the trade, not the chase.
• Watch the consume. When price eats through a zone and it goes dark, that fuel is gone. Do not expect it to defend again unless fresh leverage rebuilds there. A zone that keeps re-lighting near price is real, standing demand.
• Use the leverage toggles to pick your distance. The full map is dominated by the dense high-leverage fuel within roughly 2% of price. Filter to 25× and the deeper 2 to 4% band lights up: the zone where wider stops and lower-leverage positions actually break, which the composite view visually underweights. Scalps read the full map; swing stops read the 25× view.
• Place risk on the far side of fuel. Do not tuck a stop just beyond a fat cluster; that is exactly where price gets hunted to trigger the cascade. Put stops past the whole cluster, and take profit into a level, where the forced flow hands you exit liquidity.

Make it yours

The Settings panel on the chart tunes how the field reads, without ever changing the data underneath. Low decides how much weak fuel is culled into black: raise it for a sparser, higher-conviction map, lower it to fill in more texture. Peak sets how rare ignition yellow is. Gamma darkens or lifts the midtones. Opacity and the colormap (Ember, Inferno, Magma, Viridis) set the overall look. Same zones, same prices, same history, your read.

Why it’s different

• Universal. The field is computed from price action, so it is there on every market, including the 90% of alts with no clean open-interest feed. Most liq maps go blank exactly where the data thins.
• Replay-perfect. Because it is derived from candles, the map reproduces exactly on historical data, with no baked snapshots and no drift. Rewind any move and the fuel rebuilds the way it truly did.
• Static and trustworthy. A zone sits at its real price with the same brightness no matter how you zoom or scroll. The map never reshuffles under you, so a level you marked is still there when you come back to it.
• Located, not smeared. Fuel sits where positions actually opened and where the leverage math breaks, not spread evenly across price like a gradient.
• Reads forward, not just back. Standing fuel projects right of the live candle with length proportional to its weight. The strongest magnets literally reach furthest into the future, so the right side of your chart is a ranked target list at the exact prices that matter.
• Built for the 90% movers. The grid is relative to price, so the map stays crisp through an 8× pump and back, with no smearing at the bottom of the range and no hairlines at the top. The wilder the chart, the more you need the map to hold up.
• Consumed in the right direction. A zone burns only when price reaches it from the triggering side (long fuel below on a drop, short fuel above on a rally), so a pump no longer wipes the fuel beneath it. It carves exactly where price traded and re-lights when fresh leverage returns.
• Quiet by default. Sub-median fuel melts into a near-black toe and ignition yellow is rare by design, so when something on the map is bright, it means something.
• Confirmed, not just predicted. The Levels layer scores clusters against real open interest and real liquidations, so the high-confidence spots are marked as such.
• Direct Hyperliquid evidence. Liq Levels HL uses covered real open positions, actual leverage and exchange-reported liquidation prices, with census coverage visible instead of implied.
• Honest about the unknown. Where the data genuinely cannot tell us (cross-margin positions, thin coins), we dim it rather than fake confidence.

We scored it against reality

Every liquidation map claims its bright zones matter. We measured ours. We took two weeks of real forced liquidations from the exchange feed (67,612 events, $1.45 billion across majors, mid-caps, and fresh movers) and asked one question: when a liquidation actually fired, how bright was the map at that exact price, at that exact moment, ranked against every other price nearby?

• The brightest tenth of the map captured 21.7% of all fired liquidation dollars, more than twice its share of the chart. The two brightest tenths together took roughly 40%.
• Capture rises steadily with brightness, decile by decile. Dim zones fire rarely; bright zones fire constantly. The ramp you see is the probability you would want it to be.
• It is not just “liquidations happen near price” in disguise. A naive closest-prices-win baseline explains part of BTC and almost nothing anywhere else. On mid-caps and fresh movers the map holds its edge while proximity alone collapses to noise.
• The leverage toggles are signal, not decoration. Filtering to the lower tiers (25×) re-focuses the map on the deeper 2 to 4% band, where lower-leverage positions actually break, and that filtered view captures 2.5× its share of fired dollars. Tight stops live in the near-tier view; the stop-hunt zone lives in the 25× view.

We re-run this scoreboard as the model evolves, and the numbers gate every change: if a new weighting cannot beat the current capture, it does not ship.

Built to learn from, not just stare at

The map works in replay, and because the field is computed from price itself, replay is exact, not an approximation. A live snapshot only ever shows the current fuel; it cannot show you what a cascade actually does. Replay can.

Rewind any historical move (the flash crashes, the 90% alt squeezes, yesterday’s pump) and watch the fuel build, the zones light up, the price get pulled in, and the liquidations fire in sequence. Our guided lessons walk you through each one: how the cluster formed, why price was drawn to it, and where it was always going to break. You are not memorizing a pattern from a screenshot. You are watching the mechanism move by move on real market history, until you can read it live yourself.

Read it right

The heatmap shows you where the fuel is, not where price must go. Liquidation levels are powerful but not destiny: everyone watching the same bright zone changes the game, and a level only matters once price actually approaches it. The field is an honest estimate read from flow; the Levels are where real open interest and real liquidations confirm it. Because exchanges report only a fraction of liquidations, we calibrate on where they happen, not the dollar totals, and we always label that honestly. And where the data genuinely cannot tell us (cross-margin positions, thin coins), we dim it rather than fake confidence. Use it as a map of pressure: learned in replay, applied with a plan.