Liquidation Heatmap

The Bitcoin liquidation heatmap renders a two-dimensional density grid. The horizontal axis is calendar date across the rolling year, the vertical axis is BTC price on a logarithmic scale, and the cell colour intensity reflects the modelled USD size of expected forced-unwind flow if price traded at that cell during that day. The bright white line traces the actual daily close; a single accent dashed marker indicates today’s spot. Sparse cells appear empty; dense clusters glow.

The grid is resampled from a fine-grained upstream feed of 283 price levels across 360 daily columns into 20 log-spaced price buckets × daily columns so the shipped JSON stays under 100 KB. The compression keeps zoom and pan responsive without losing the cell-level shape; hover any cell for the exact cluster size and column total. The grid refreshes overnight; spot in the reading row above auto-refreshes a few times a day in the browser.

No exchange publishes individual position liquidation prices. Every liquidation heatmap on the public web is therefore a modelled estimate — built from the open-interest snapshot, leverage-tier assumptions, and the standard maintenance-margin formulas each venue uses. Provenance is documented on the data sources page; the resampling and signal derivation are spelled out on methodology. The clearest published primary description sits in a cell-construction methodology note: the heatmap calculates the liquidation levels based on market data and different leverage amounts. The calculated levels are then added to a price bucket on the chart.

The ratio in the reading row is computed as:

cluster_up = Σp ∈ (spot, spot × 1.05] grid(today, p)
cluster_down = Σp ∈ [spot × 0.95, spot) grid(today, p)
ratio = cluster_up / cluster_down

Above 2.0, the upside cluster dominates — a small spot rally would cascade through the stacked stops into forced buy-backs (a short-squeeze setup). Below 0.5, the downside cluster dominates and a small spot drop can cascade into forced sells (a long-liquidation setup). The threshold values are descriptive cuts on the empirical distribution, not bright-line laws.

The single most important caveat is published in the same note: the Liquidation Heatmap predicts where liquidation levels are opening but not closing. Thus, the actual number of liquidations will be lower. Realised liquidations are bounded above by the modelled clusters; the chart never under-counts what would unwind, but it routinely over-counts, because positions close before they liquidate as traders cut losses or reposition. Treat dense cells as upper bounds on potential flow, not as forecasts of realised flow.

Three regimes resolve from the ±5% cluster ratio. The numerical thresholds have weakened over time as the rolling window’s composition shifts; treat them as anchors on the recent distribution, not absolute laws. A ratio above 2.0 historically precedes upside flow; a ratio below 0.5 precedes downside flow; the middle band is a coin-flip from the heatmap alone, and other indicators have to break the tie.

Seven monthly snapshots through the rolling year sketch the regime rotation. Each row reports the ±5% cluster ratio on the snapshot day, the spot price the cluster centred on, and the regime that ratio resolved into. Cells where one side of the cluster goes to zero (no meaningful leverage stacked on that side of spot) are flagged as such directly; the ratio in those cases has no finite value but directionally clean.

The folkloric framing is that price “seeks” dense liquidation clusters because market makers know where forced flow will hit and probe those zones to trigger it. There is microstructure substance to that claim — the venues that move spot also see the modelled cluster maps, and the realised path of price in many windows tracks the published magnet zones closely. The cleanest recent positive case is spring 2024, when sub-$50k upside clusters magnetised wicks during the early-cycle consolidation before BTC extended higher.

The pattern is most reliable in low-dispersion windows — when spot is grinding, funding is near the structural baseline, and open interest is range-bound. In those regimes, the cluster geometry is the dominant microstructure feature, and price probes the dense cells before reversing. The published heatmap captures that geometry honestly. Treat it as a useful prior on where price may probe, not as a price target.

The cleanest counter-example is November 2022. In the days before FTX’s Chapter 11 filing, the dominant cluster on every published heatmap sat on FTX-perp positions below the prevailing spot. Price did not seek that magnet on the way down — the magnet evaporated. Positions were socialised into the bankruptcy estate; the cluster zeroed out by mid-November not because price obeyed it but because the venue that carried it disappeared. The post-FTX market-structure analysis put FTX’s peak derivatives market share at roughly 15%; that share didn’t drift, it disappeared in one weekend, and the heatmap of November 7 documented a regime that no longer existed by November 14.

The honest framing: the heatmap is a topographical map of leverage as it currently sits, conditional on the venues currently in the model staying live. When that conditional breaks — venue failure, regulatory shutdown, mass migration — the magnet narrative breaks with it. The chart is documenting a regime that may not survive the day.

For a dollar-cost-averaging investor. Functionally none. The heatmap operates on day-to-week horizons and resolves to spot price probes that mean little over a multi-cycle DCA. Skim the regime row for cycle texture; otherwise this chart is not for you.

For a cycle-timing trader. Treat dense clusters as conditional liquidity, not as price targets. The cluster tells you that if price reaches that bucket, leverage will unwind there; it does not tell you whether price will. Pair with open interest — the more total leverage stacked on the side opposite spot, the sharper any cascade — and with funding rate to read whether the crowded side is paying real carry to stay in.

For a researcher. The grid, the price-bucket centres, the daily timestamps, and the ±5% cluster derivation are the only inputs. The methodology page documents the resampling, the bucket spacing, and the regime threshold derivation; the canonical methodology page is the cleanest primary description of the cell-construction model.

Modelled, not measured. Per the canonical methodology note, the heatmap predicts where liquidation levels are opening but not closing. Thus, the actual number of liquidations will be lower. Realised liquidations depend on real-time balances, leverage choices that change continuously, and venue-specific margin rules. Dense cells are upper bounds on potential flow, not forecasts of realised flow.

Venue discontinuity. November 2022 is the canonical failure mode — the dominant cluster vanished overnight when the venue carrying it failed. Any heatmap is conditional on the model’s venue list staying live; a regulatory shutdown, a delisting, or a venue-failure event can erase a cluster without any trader closing a position. Treat dense clusters during dislocation windows with extreme scepticism.

Tactical horizon. The heatmap is a day-to-week tool, not a cycle-level lens. It says nothing about whether Bitcoin is rich or cheap on a multi-year horizon — that work belongs to the realised-price, RHODL, and NUPL charts. A reader reaching the heatmap from a multi-month framing will over-interpret short-term cluster structure as signal. The right response is to walk back to a longer indicator and let the heatmap inform the entry geometry, not the call.

Canonical questions readers bring to liquidation heatmaps, answered against the rolling-year grid powering this page.