FuturesAI Liquidation Map: How to Test Its 78% Claim
Quick answer
FuturesAI’s liquidation heatmap article, dated March 3, 2026, says 78% of large liquidation cascades keep moving in their direction. It gives no sample size and no time period. A 78% rate from 10 cascades could really be anywhere from 47% to 93%. Its own map tables also fail to add up to its headline totals. Treat the claim as a hypothesis, and size any trade as if it were half as strong.
A “78% win rate” with no count behind it is a story, not a statistic.
FuturesAI publishes a liquidation heatmap article with a map of BTC liquidation clusters and a cascade claim. This post recomputes what can be recomputed and shows what the number would have to be for a trade to pay. It is arithmetic on a public page, not a signal, and nothing here is investment advice. Leverage can lose your full principal.
Jump to a Section
- What does the FuturesAI liquidation article claim?
- Do the map tables add up to the headline totals?
- How reliable is a 78% rate with no sample size?
- How far is price from the named trigger zones?
- What loss size makes a 78% cascade trade break even?
- What to Ask Before You Trade a Published Win Rate
- Tips
- Warnings
- FAQ
What does the FuturesAI liquidation article claim?
The FuturesAI liquidation heatmap article, dated March 3, 2026, reports $2.7B of long liquidations below the BTC price and $4.6B of short liquidations above it, defines a cascade as more than $500M liquidated in one hour, and says cascades continued in their direction 78% of the time with a mean move of 4.2% from trigger to exhaustion (futuresai.io).
| Claim on the page | Figure | Sample size or period given? |
|---|---|---|
| Long liquidations below price | $2.7B | No |
| Short liquidations above price | $4.6B | No |
| Cascade threshold | $500M in 1 hour | Definition |
| Continuation in cascade direction | 78% | No |
| Mean cascade move | 4.2% | No |
| Retrace 50% within 48 hours | 67% | No |
When I checked the page for a sample size or a backtest window, I found that neither is stated anywhere. That is the main finding, and the rest of this post works around it.
Do the map tables add up to the headline totals?
The FuturesAI liquidation map table does not add up to its own headline: the long rows sum to $2,992M against a stated $2.7B, and the short rows sum to $4,810M against a stated $4.6B.
| Side | Rows in the table | Sum of rows | Headline | Headline as % of rows |
|---|---|---|---|---|
| Long | $842M + $1,240M + $620M + $290M | $2,992M | $2.7B | about 90.2% |
| Short | $180M + $950M + $1,580M + $2,100M | $4,810M | $4.6B | about 95.6% |
A gap of 4% to 10% is not dramatic. It could be rounding or a different cutoff. But the page does not explain it, so I would not size anything to the nearest million.

How reliable is a 78% rate with no sample size?
A 78% continuation rate measured on only 10 cascades has a 95% Wilson confidence interval of roughly 47% to 93%, and even 100 cascades leaves it between about 69% and 85%, so the sample size matters more than the headline percentage.
| If the 78% came from this many cascades | 95% interval for the true rate |
|---|---|
| 10 | 47.1% to 93.4% |
| 25 | 58.7% to 89.8% |
| 50 | 64.8% to 87.2% |
| 100 | 68.9% to 85.0% |
| 500 | 74.2% to 81.4% |
I computed those with the Wilson score formula. They describe sampling noise only. They do not cover a pattern that stopped working, or cherry-picked dates.
How far is price from the named trigger zones?
With BTC at the page’s $64,200, the named long trigger at $60,500 is 5.76% below and the named short trigger at $70,000 is 9.03% above, so a 4.2% mean cascade move from the lower trigger would land near $57,959, almost exactly on the page’s $58,000 cluster of $842M.
That is a tidy story. It is also exactly what you would expect from a map drawn after the fact. Here’s where people get tripped up: a cluster that price tends to reach is not the same as a cluster you can trade profitably.
What loss size makes a 78% cascade trade break even?
A trade that gains the page’s 4.2% mean move on 78% of cascades only breaks even if the average loss on the other 22% stays under about 14.9% of price, because 0.78 ร 4.2% = 3.28%, and 3.28% รท 0.22 = 14.9% (before fees, with no leverage).
| True continuation rate | Average gain | Average loss that wipes it out |
|---|---|---|
| 78% (page’s claim) | 4.2% | 14.9% |
| 69% (lower bound if n = 100) | 4.2% | 9.3% |
| 60% | 4.2% | 6.3% |
| 55% | 4.2% | 5.1% |
Leverage changes the picture. At 10x, a 10% adverse move erases the margin, so the loss side is capped at 100% of margin while the 4.2% gain becomes 42% of margin. Fees then take a fixed bite of notional. These are illustrations of the arithmetic, not a forecast, and the page’s own mean move may not describe the typical cascade.
Run your own version at FuturesAI Payback
What to Ask Before You Trade a Published Win Rate
Before you trade any published win rate, find out how many trades produced it, over what dates, whether fees and slippage were deducted, and how large the average loss was compared with the average gain.
- How many events? A rate with no count is unverifiable.
- What period? A rate from one bull run says little about a range market.
- What is the average loss? A high win rate with large losses can still lose money.
- Were fees included? Round-trip fees are paid on every trade.
For another worked check of a published number, see FuturesAI BTC Funding Signal: How to Check Its 68.7% Claim. For streak math, see FuturesAI Hit Rate: How Long a Losing Streak Gets.
Tips
- Write down the sample size next to every win rate you collect. If there is none, mark it “unverified”.
- Use the lower bound of the interval for sizing, not the headline.
- Pair any win rate with an average loss before you judge it.
- Re-add the table rows yourself. Mismatched totals are a prompt to ask questions.
Warnings
- This is not investment advice. The page’s statistics may not repeat.
- Leverage can lose your full principal, and liquidation can come before your stop.
- The article gives no backtest window, so past performance is unknown.
- Do not size a trade to a number the source cannot reconcile.
FAQ
What is a liquidation cascade?
FuturesAI’s article defines a cascade as more than $500M of liquidations within one hour. Forced closes push price further, which triggers more forced closes. It describes cascade size, not a guaranteed direction.
Is the 78% cascade win rate reliable?
It cannot be verified from the page, because no sample size or date range is given. If it came from 100 events, the 95% interval is about 69% to 85%. With 10 events it is about 47% to 93%.
Can a liquidation heatmap predict where BTC will go?
It shows where leveraged positions would be forced out, not where price will move. A cluster may attract price or may be avoided. Treat it as context, and test any rule with fees and a known sample.
Sources
๐ Related guides
๐ Hub guide: For the full earn-save-flip playbook โ pricing, taxes, cashback, and collectible margins โ see the Income & Smart Shopping Hub.
Written by Jordan Lee. Figures recomputed from FuturesAI’s published page as of November 8, 2026. Not investment advice; leverage can lose the full principal.
