TL;DR. The liquidation map on Coinglass is a model, not a live order book. It estimates where leveraged positions will hit forced-close prices by combining the public liquidation stream, posted leverage tiers and OI distribution. Liquidation clusters are magnets for price β not because the chart predicts the future but because real forced flow accelerates near concentrated stops. Used with funding and OI it helps flag where an ordinary pullback could turn into a cascade.
1. What the liquidation map actually is.
Open the CoinGlass Liquidation Heatmap dashboard and you see horizontal price bands shaded by estimated liquidation depth. Bright clusters above the current price are short-side stop concentrations; clusters below are long-side stops. The map is model output, not a real-time order book β Coinglass and similar venues do not have access to every exchange's positions, so they infer cluster locations from three public signals.
2. Where the data comes from.
(1) The public liquidation stream. Binance and other large venues stream every individual liquidation in real time (Binance Futures WebSocket exposes the !forceOrder feed). Coinglass and Kaiko aggregate this. (2) Public leverage tiers. Each venue posts maximum-leverage tables by symbol and notional size; Binance's table is public on its Futures contract specs page. (3) OI distribution. Coinglass and other dashboards report exchange-level OI by symbol. Stitching those three together, Coinglass back-solves probable liquidation prices and aggregates them into the heatmap. The result is an estimate, sometimes off by 5β15% depending on regime β useful as a regional indicator, not a precise price target.
3. Why the "$4 billion liquidated" headlines are inflated.
Three reasons. Notional vs realized loss: the headline number is the dollar notional of the liquidated position, not the trader's actual loss. On 10Γ leverage, a $10,000 notional liquidation cost the trader $1,000. Push rate limits and amplification: Binance throttles its public WebSocket force-order stream (one message per second per symbol during cascades), so trackers extrapolate from sampled events, and headline liquidation totals are estimates rather than exact counts.Cross-venue double-counting: aggregators sometimes count the same trader's hedged position on multiple venues twice. The realistic correction: discount published headline numbers by 30β60% when sizing positions against them.
4. Liquidation pools tell you where price wants to go.
Large clusters are price magnets. Mechanism: as price approaches the cluster, leveraged positions start to hit margin calls; the resulting forced-close orders accelerate the move into the cluster; once cleared, the move loses its fuel and reverses (background: Binance Academy: how crypto liquidation works). The practical reads: thick liquidation pool below current price = short-term downside target; thick pool above = upside target; symmetric pools = whippy chop until one side resolves. Stack with the open interest guide β high OI + a clustered liquidation map = cascade-prone environment.
How much weight does the magnet idea deserve on the coin you trade? Rather than trusting anyone's published "hit rate", you can measure it yourself:
- At a fixed time each day, screenshot the liquidation map and note the thickest cluster above and below price, plus its distance from price (bucket it: within Β±1%, Β±1β3%, Β±3β5%, beyond Β±5%).
- Twenty-four hours later, check the Binance candle high and low: did price reach that cluster?
- Once you have a few dozen samples, work out the touch rate per bucket, and compare it with random price levels at the same distance.
You will most likely find that near clusters get touched far more often than distant ones, and that a good part of that is simply normal price movement rather than any "magnet". So treat a thick cluster close to price as a real short-term reference, and do not treat clusters beyond Β±5% as imminent targets.
5. Three real cases β Aug 2024, Nov 2024 breakout, Feb 2025 alts.
Binance BTC/USDT 2024-08-05: open $58,161, intraday low $49,000 (β15.8%), close $54,018. According to CoinDesk's same-day report, liquidations across crypto futures topped $1 billion over 24 hours, and roughly 87% of the traders liquidated were long. Price fell through one level after another as the yen carry unwind cascaded into crypto (background: VOA: Japan stocks plunge, leading global market sell-off). The general lesson still holds: the map did not predict the trigger (the macro flow did). At best it shows where forced selling is likely to speed up once price gets there, and a low that lands near a big cluster is something you notice afterwards, not a level you can reliably trade in advance.
Binance BTC/USDT closed 2024-11-05 at $69,372. On 11-06, price first traded above $73,000 around 02:00 UTC, cleared the March high of $73,777 within the next hour, and reached $76,400 before the day was out (daily close $75,572). What the map is useful for in a move like this: short stops and short liquidation levels tend to stack just above a prior all-time high, and a clean break can turn them into fuel for a fast extension. Read ahead of time, the map tells you where acceleration is likely if the level breaks; it cannot tell you whether it will break.
On February 3, 2025, Binance daily candles show BTC falling 6.6% from the open to the intraday low, while SOL dropped 13.6%, DOGE 24.7% and SUI 27.7% over the same open-to-low stretch. Alts falling two to four times as far as BTC is what crowded leveraged longs sitting closer to their liquidation prices looks like. The single most actionable takeaway: during macro shocks, altcoins with the thickest long clusters get punished hardest. The reverse is true at relief bounces β short-side cluster density above current price predicts which alts squeeze hardest.
6. Four common misreadings.
"Price always hits the cluster." No. The cluster is gravitational only when supporting catalysts exist β a 1% pullback into a small cluster usually dissolves it without much price action. Real magnetism requires either a macro flush or a coordinated leverage build-up nearby.
"Big liquidation total = time to buy the dip." Headline liquidation numbers are notional, inflated and lagging. A "billion-dollar liquidation day" reported by Cointelegraph might represent $300β500M in actual losses spread across thousands of traders β not the structural capitulation it sounds like. Watch the funding response: real bottoms come with sustained negative funding 24h+, not single-print liquidation spikes.
"The liquidation map is a real-time order book." It is a back-solved model from public data, refreshed every few minutes. Treat the cluster prices as Β±5β15% probabilistic regions, not precise levels.
"Set your stop just below the cluster β you'll be safe." Worst trade in the playbook. Smart-money knows where retail puts stops; placing yours just outside the cluster guarantees you get swept first. Stops belong at structural levels (prior swing low, key moving average), not at cluster edges.
7. Liquidation map + OI + funding β three-way stack.
The strongest cascade-warning configuration: OI at all-time-high, funding rate above P90 sustained 24h+, and a thick liquidation cluster within 5β7% of current price. Any minor catalyst at this density triggers the cascade. Inverse for bottoms: OI down, funding negative 24h+, dense short cluster above β a relief squeeze is loaded. Read this together with funding rate and open interest.
The four configurations below are the ones to memorize β they map a three-layer signal into a single read with high consistency.
| Open interest | Funding | Liquidation map | Combined read |
|---|---|---|---|
| All-time high | Positive P90+ | Downside cluster thicker than upside Γ2 | Cascade-liquidation danger zone |
| All-time high | Negative P10- | Upside cluster thicker than downside Γ2 | Short-squeeze rebound setup |
| Clearly off highs | Near zero | Both sides thin | Consolidation phase, wait for direction |
| Rising | P75 positive | Both sides thick | Pre-breakout setup, volatility amplification incoming |
Coinglass and Coinalyze both publish the underlying OI, funding and liquidation data; the table above maps those readings to leverage conditions. Note that the second row (the short-squeeze setup) is statistically the highest-conviction long entry pattern in this framework β it has preceded most of the major relief rallies of 2023-2025.
8. FAQ.
Is the liquidation map predictive or descriptive?
Descriptive β it shows where current leverage sits. It becomes predictive only when overlaid with a catalyst (macro event, breakout, ETF flow shift).
How accurate are the cluster prices?
Within Β±5β15% depending on the leverage-tier model used. Treat them as regions, not precise levels.
Why is the headline liquidation number inflated?
Three reasons: notional vs realized loss, WebSocket push-rate-limit scaling factors, and cross-venue double counting. Discount headlines by 30β60% before sizing decisions against them.
Should I place stops just outside a cluster?
No. Clusters get swept first. Place stops at price-structure levels β prior swing low, key MA, value-area edge β not at cluster boundaries.
When does the liquidation map matter most?
When stacked with OI ATH and funding at P90+. That combination identifies cascade-prone regimes where a small move triggers an outsized one.
What to read next.
Crypto assets are volatile and not suitable for every investor. This page is editorial analysis, not financial advice.