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.

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 directly; the ratio in those cases has no finite value but is 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 heatmap’s structural weakness is that the thing it maps — open leverage — can change shape without a single trade printing at the bucket it labels. A cluster can evaporate, and when it does the magnet narrative fails not gradually but all at once. Three mechanisms do this, and all three have specific, dateable precedents.

Venue failure (the FTX case, November 2022). In the days before FTX’s Chapter 11 filing on 11 November 2022, 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 7 November documented a regime that no longer existed by 14 November.

Funding-driven flush before the stops are reached. When one side of the book is crowded and paying steep funding — as longs were into the 17 April 2021 and the late-cycle 2024 funding spikes — positions get cut for carry, not for margin. The crowded cluster thins from the top down as traders pay to exit, so the dense cell the map advertised is smaller by the time price arrives, and the cascade it implied never fully fires. Funding is the single best confound to cross-read: a lopsided cluster that the crowded side is bleeding to hold is a cluster on borrowed time.

Venue-mix and custody drift in the model’s denominator. The map is only as honest as its venue list. As open interest has migrated — Binance’s share compressing post-2023 as CME and regulated perp venues grew, and again as ETF-era basis trades parked offsetting futures against spot — a rising fraction of the “real” leveraged book is either off the model or is delta-hedged and will never liquidate the way a naked directional position would. The cluster you read is conditional on the venues currently in the model staying live and staying representative; when the mix shifts, the absolute USD figures drift even when the underlying risk has not.

The horizon mismatch. Even when none of the above breaks it, the heatmap is a day-to-week tool, not a cycle 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 arriving from a multi-month framing will over-read short-term cluster structure as signal. Walk back to a longer indicator for the call, and let the heatmap inform only the entry geometry.

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 it 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. A one-sided cluster that the crowd is bleeding to hold is the highest-quality version of this setup; a one-sided cluster with flat funding is far weaker.

For longer-horizon readers the chart asks almost nothing. It resolves to spot-price probes that mean little over a multi-cycle hold — skim the regime row for near-term texture and otherwise read the cycle off slower instruments. The one durable use across horizons is the failure case above: when a dominant cluster sits on a single venue or a single crowded, funding-paying side, discount it, because that is exactly the cluster most likely to vanish before price ever tests it.