CoinGlass liquidation heatmap vs MarketTrace: the same question, two different answers
What the CoinGlass liquidation heatmap shows, what it costs, and how MarketTrace's free map differs: expected and realized layers on one canvas.
Both pages answer the same question: where would open positions on BTC perpetuals get force-closed if price moved there. CoinGlass answers it with a predicted map: three models, windows from 12 hours to two years, 20 exchanges, with only the default 24-hour view open without an account and the full version part of a $28-a-month Prime plan. MarketTrace answers it with a free 24-hour map that draws the prediction and the liquidations that actually printed on one canvas, publishes the model's rules, and checks the prediction against the tape every hour. Which one you want depends on whether you need history or a check.
Everything below was read from the two sites on 23 October 2026, without logging in. Prices and gates change; the date is there so you can re-check. The MarketTrace BTC liquidation heatmap is open in another tab if you want to follow along on the free side.
What each page draws
CoinGlass describes its heatmap in its own words: "In short, it predicts the price levels at which large-scale liquidation events may occur." The levels are computed "based on market data and different leverage amounts", added up by price, and coloured on one scale: "The color range goes from purple to yellow, with yellow representing a high number of predicted liquidation levels." That is one layer, the prediction, and CoinGlass is careful about what it is. The page says the map "predicts where liquidation levels are likely to initiate but not where they will stop. Therefore, the actual number of liquidations will be fewer", and that the size "must be viewed as a relative figure by comparing it with other levels". Realized liquidations exist on CoinGlass too, on separate pages: Liquidation History, Top Liquidation Events, and a Liquidation History Heatmap.
MarketTrace draws two layers on the same canvas. The expected layer is a model, liqmap_model_v1, with five rules published on the methodology page. Open-interest increases become lots. Lots are split long and short by the taker buy ratio, then into equal thirds at 25x, 50x and 100x, and each third gets a level from the venues' maintenance-margin rates. Falling open interest shrinks every lot. A level that price crosses leaves the map. The realized layer is the liquidation tape from Binance, Bybit and OKX, placed at the trade price where one is known. Minutes whose liquidations were large against the order book within 3 bps of price get rings; a profile by price sits on the right and an hourly tape underneath. The third element is the hit-rate: every hour, each expected level that price crossed is checked against the tape for same-side prints within one step and 0.3% of the level. The number is published once there are 90 days of tape, under 1% feed gaps and 30 days on one model version; until then the page shows a day counter.
Side by side
| CoinGlass Liquidation Heatmap | MarketTrace Liquidation Heatmap | |
|---|---|---|
| What is drawn | Predicted liquidation levels on one intensity scale, purple to yellow | Expected levels from a model and the liquidations that printed, on one canvas |
| Executed liquidations on the chart | Not on this chart; separate pages: Liquidation History, Top Liquidation Events, Liquidation History Heatmap | Yes: Binance, Bybit and OKX, at the trade price where one is known, with a price profile, an hourly tape and rings on the minutes whose liquidations were large against the order book |
| Model | Three models (Model 1, 2 and 3); the method is described in general terms, "based on market data and different leverage amounts" | One model, liqmap_model_v1: five rules published in full, the version on every response |
| Leverage assumption | Not disclosed; the heatmap page has no leverage selector | 25x, 50x and 100x in equal parts, stated on the page; each leverage as its own layer is planned for Pro, not on sale yet |
| Accuracy record | Not stated; the page calls the size "a relative figure" and says "the actual number of liquidations will be fewer" | Hit-rate against the liquidation tape, computed every hour, published after at least 90 days of tape, with under 1% feed gaps and 30 days on one model version |
| Coverage limits | Not stated | Shown above the chart: Binance's count is a floor (one print per second), OKX is a sample, Hyperliquid has no feed |
| Exchanges | One pair at a time: 42 pairs in the BTC list on 20 exchanges, Hyperliquid among them; or a coin-level view on the Symbol tab | Binance, Bybit and OKX together for both layers; Hyperliquid has no liquidation feed and is left out, and the page says so |
| Time windows | 12 hours, 24 hours, 48 hours, 3 days, 1 and 2 weeks, 1, 3 and 6 months, 1 and 2 years | 24 hours; 7 and 30 days in Pro, not on sale yet |
| Without an account | The default 24-hour Binance BTC/USDT view; every other window, 12 hours included, opened a log-in prompt | The whole page: the 24h map for 9 assets, both layers, the levels table, the cascade list and the method |
| Assets | Many coins and pairs; Prime opens the heatmap for all coins | 9: BTC, ETH, SOL, BNB, XRP, DOGE, HYPE, ZEC, ENA |
| Refresh | Automatic refresh is part of Prime; without it the page is refreshed by hand | Every 60 seconds, no account |
| Price | Prime: $28 a month, $78 a quarter or $268 a year | 24h map free; Pro planned at $4.99 a month or $49 a year, not on sale yet |
| Heatmap data by API | Professional API plan or higher: $699 a month, or $8,388 a year | The 24h inputs and the liquidation tape as public JSON with no key, and a free MCP server |
Checked 23 October 2026. The CoinGlass column comes from CoinGlass's own pages, opened without an account and listed under Sources; access can differ by region or account.
Two rows deserve a note. Without an account, every CoinGlass window other than the default 24 hours, 12 hours included, opened a "Log in to unlock full data" prompt for us; the Prime page adds that non-Prime accounts get the 6- and 12-month views for BTC and ETH only. And the exchange count is the BTC pair list on the heatmap page: 42 pairs on 20 exchanges.
When CoinGlass fits better
History. If you want to see where the bands sat in March, or trace a year of clusters above and below a range, CoinGlass has the windows and MarketTrace does not. Breadth. CoinGlass covers far more pairs and venues, Hyperliquid among them, and has a separate Hyperliquid liquidation map; MarketTrace leaves Hyperliquid out of its model because Hyperliquid publishes no liquidation feed. Choice of model. Three models mean three opinions about the same open interest, for readers who like to triangulate; full access to all three is part of Prime. And the page is established: many of the screenshots that circulate in trading chats and in posts explaining liquidation heatmaps are CoinGlass screenshots, so its colour scale is the one people already read.
When MarketTrace fits better
The check. A predicted map on its own cannot tell you whether it was right last Tuesday. MarketTrace puts the tape on the same canvas, so when price crosses a band you see whether prints appeared under it, and the hit-rate turns that into a number once there are at least 90 days of tape, under 1% feed gaps and 30 days on one model version. The method. The five rules of the model are public, including the leverage split and the margin rates, so you know what a band is made of and can disagree with it. The coverage line. Every map built on public liquidation feeds inherits the Binance one-per-second snapshot and the OKX sample; MarketTrace prints that above the chart instead of leaving it in a footnote. And the price. The 24-hour map, both layers, the levels table, the cascade list and the methodology are free without an account, and the same tape is queryable by agents through the free MCP server.
The disagreement that matters
The two pages disagree about what the glow means. CoinGlass offers a reading: "Magnet Zone: Concentration of potential liquidation levels within a specific price range may indicate that the price is likely to move towards that zone." The MarketTrace map does not call any level a magnet, and will not until the hit-rate says something. Our position, written up in how accurate are liquidation heatmaps, is that the magnet claim is a hypothesis that deserves a number, and that the number has to come from comparing expected levels with the realized tape, level by level. The two pages are built on different ideas of what a reader should be given: an interpretation or a measurement. Pick the one you want to be handed.
If you are new to the picture itself, the reading guide is here: how to read a liquidation heatmap. The order-book side, which is a different map altogether, lives on the depth page.
FAQ
Is the CoinGlass liquidation heatmap free?
Only the default view. The 24-hour Binance BTC/USDT map loaded for us without an account on 23 October 2026; every other window, 12 hours included, opened a "Log in to unlock full data" prompt. CoinGlass's Prime page says the 6- and 12-month views are limited to BTC and ETH without Prime, and that Prime ($28 a month, $268 a year) adds all coins, 24-month+ ranges and automatic refresh.
Is the MarketTrace liquidation heatmap free?
Yes. The 24-hour map with both layers, the table of expected levels, the recent cascades and the methodology are free with no login. 7-day and 30-day windows are planned for a Pro tier that is not on sale yet.
Do the two maps show the same levels?
No, and they should not be expected to. Each is a model on an assumed leverage mix and a margin table, and the two models are different. The shapes tend to rhyme because both start from open interest; the distances and intensities differ. Neither is a record of where real liquidations are waiting, because no exchange publishes that.
Which one is more accurate?
Nobody can say yet. CoinGlass does not publish a hit-rate, and MarketTrace's is published only once there are 90 days of tape, under 1% feed gaps and 30 days on one model version. What you can compare today is what each page tells you about its own limits: CoinGlass says the actual number of liquidations "will be fewer" than shown; MarketTrace prints the feed caps above the chart and shows the realized prints next to the bands.
Does either map cover Hyperliquid?
CoinGlass lists Hyperliquid BTC/USD among its heatmap pairs and has a separate Hyperliquid Liquidation Map page. MarketTrace leaves Hyperliquid out of the model and the tape because it has no liquidation feed, and says so in the coverage line; at the time of writing Hyperliquid holds 19% of BTC open interest.
Can I get the data through an API?
CoinGlass sells API access on separate plans from $29 a month; its liquidation heatmap endpoint is on the Professional plan ($699 a month) and above. MarketTrace's liquidation archive is open at /api/liquidations/range, and the MCP server that wraps it is free.
Sources
- CoinGlass, Liquidation Heatmap: page text, model tabs, window and pair selectors, read 23 Oct 2026 without an account.
- CoinGlass, Prime: plans and what they add to the heatmap, read 23 Oct 2026.
- CoinGlass, API pricing, read 23 Oct 2026.
- CoinGlass, Pair Liquidation Heatmap API documentation: the plans that include it, read 23 Oct 2026.
- MarketTrace, BTC liquidation heatmap and liquidations methodology, read 23 Oct 2026.
Descriptive microstructure, not financial advice. CoinGlass is a trademark of its owner; it is named here to compare two products.