FuturesAI BTC Funding Signal: How to Check Its 68.7% Claim
FuturesAI publishes research notes under its Insights tab, and one of them says negative Bitcoin funding has preceded rallies. It comes with a snapshot table, a win rate and a median forward return.
Instead of repeating the headline, I wanted to run the page’s own numbers against each other. Three checks, all done with a calculator.
Jump to a Section
- Can 847 instances happen 3.2 times a year?
- Why does the page’s annualized funding look too high?
- What does an 8.2% drawdown do to a leveraged position?
- How much does negative funding pay a long against fees?
- A Four-Line Audit for Any Research Note
Can 847 instances happen 3.2 times a year?
FuturesAI’s BTC funding rate page, dated March 8, 2026, reports 847 instances where the 7-day average funding rate fell below -0.01% across major exchanges, and also says the signal occurs about 3.2 times per year; 847 divided by 3.2 is about 265 years. Bitcoin perpetual swaps have existed since 2016, so roughly a decade of history at 3.2 a year would give about 32 independent events.
The likeliest explanation is overlap. If every day inside a negative stretch counts as a fresh instance, 847 instances can come from far fewer episodes. The page does not say so.
That matters because the win rate depends on it. A 68.7% hit rate on 847 independent cases has a 95% Wilson range of 65.5% to 71.7%. The same rate on 32 episodes is about 22 of 32, with a range of 51.4% to 82.0%. When I compared the two ranges, I noticed that the second one starts barely above a coin flip.
| Reading of the sample | Hits / cases | Hit rate | 95% range (my calculation) |
|---|---|---|---|
| 847 independent instances (as printed) | 582 / 847 | 68.7% | 65.5% to 71.7% |
| About 32 episodes (3.2 a year x 10 years) | 22 / 32 | 68.8% | 51.4% to 82.0% |
| Half that, 16 episodes | 11 / 16 | 68.8% | 44.4% to 85.8% |
Source for 68.7%, 847 and 3.2 a year: FuturesAI insight page “BTC Funding Rate Analysis: What Negative Rates Signal” (dated March 8, 2026). Hit counts are rounded; ranges are Wilson 95% intervals computed by me.
Why does the page’s annualized funding look too high?
An 8-hour funding rate annualizes as the rate times 3 payments a day times 365 days, a factor of 1,095, so a -0.0120% rate is about -13.14% a year, but FuturesAI’s page prints -15.77% for Binance. Every row of its snapshot table is exactly 1.2 times the standard result.
| Exchange | BTC funding (8h) on page | Annualized on page | My recalculation (x1,095) |
|---|---|---|---|
| Binance | -0.0120% | -15.77% | -13.14% |
| Bybit | -0.0095% | -12.48% | -10.40% |
| OKX | -0.0105% | -13.80% | -11.50% |
| Bitget | -0.0088% | -11.56% | -9.64% |
| dYdX | -0.0132% | -17.35% | -14.45% |
Maybe the page uses a different compounding or day count. It doesn’t say. Either way, I found that rebuilding the column from the 8-hour rate takes thirty seconds and closes a 20% gap.

What does an 8.2% drawdown do to a leveraged position?
A leveraged long loses roughly its leverage multiple times the price drop as a share of margin, so the page’s 8.2% median drawdown before recovery costs about 41% of margin at 5x, 82% at 10x, and wipes out a position at 12x or higher. This simple version ignores maintenance margin and fees, which make real liquidation come sooner.
| Leverage | Adverse move that erases margin (approx.) | Margin lost at an 8.2% drawdown |
|---|---|---|
| 5x | 20.0% | 41% |
| 10x | 10.0% | 82% |
| 12x | 8.3% | 98% |
| 15x | 6.7% | liquidated (123%) |
The drawdown figure is a median. Half of the page’s cases went deeper than 8.2% before they recovered. A 68.7% chance of a positive 30-day return is no help to a trader who was liquidated on day four.
Honestly, this is the check that matters most. A win rate and a survivable path are different things.
How much does negative funding pay a long against fees?
On a $10,000 long, an average funding rate of -0.0108% per 8 hours pays the long about $1.08 per payment, or about $35.64 over the 11 consecutive days the page cites (33 payments). The page’s 7-day average line carries no interval label, so I assume it is per 8 hours like its snapshot table.
Now set that against fees. FuturesAI’s payback page lists Bitget at 0.06% market and 0.02% limit, with effective rates of 0.015% and 0.005% after discount and payback for a new account.
- Round trip at the listed Bitget market rate: $10,000 x 0.06% x 2 = $12.00.
- Round trip at the page’s effective Bitget market rate: $10,000 x 0.015% x 2 = $3.00.
- Funding received over 11 days, on the page’s assumption: about $35.64.
So funding covers the fees many times over, if the rate holds and price does not move against you. A 1% adverse move on $10,000 is already $100. Funding income is the small number in this trade, and the price path is the big one.
For the other side of this math, I walked through how long a hold takes to erase payback in FuturesAI Payback vs Funding: How Long a Hold Erases It.
A Four-Line Audit for Any Research Note
- Divide the sample count by the stated frequency. If the years are absurd, the samples overlap.
- Rebuild one annualized figure from its raw rate yourself.
- Pair every win rate with the drawdown it came with, then divide by your leverage.
- Check the date. A market snapshot is only as current as its publication day.
Full sample-size rules for hit rates are in FuturesAI Signals: A 3-Gate Filter Before You Trust One.
Tips
- Pull live funding from the exchange itself before comparing with any March snapshot.
- Write your own annualization formula in a spreadsheet once and reuse it.
- Size so the page’s median drawdown cannot liquidate you, then add a buffer.
Warnings
- Funding can stay negative for weeks while price keeps falling. Collecting it does not protect the position.
- The calculations here ignore maintenance margin, trading fees beyond those named, and funding changes between payments.
- This analysis audits the arithmetic of a published note. It is not a prediction and not advice to buy or sell. Leveraged futures can lose the full principal, and past patterns do not guarantee future results.
FAQ
Does a 68.7% win rate on a funding-rate signal mean I will win 68.7% of trades?
No. The figure on FuturesAI’s BTC funding rate insight page describes how often the 30-day forward return was positive in past instances, not how often a leveraged trade would survive its drawdown. The same page lists a median drawdown of 8.2% before recovery, which would liquidate a position at 12x leverage or higher.
How do I annualize an 8-hour funding rate?
Multiply the 8-hour rate by 3 payments a day and by 365 days. A -0.0120% rate becomes about -13.14% a year, while FuturesAI’s page prints -15.77% for the same figure, a 1.2x gap that its article does not explain.
Is a funding-rate signal from March 2026 still current in November 2026?
No. The FuturesAI insight page is dated March 8, 2026, and its snapshot of exchange rates and its 11-day negative-funding streak describe that month. Check live funding on the exchange before using any of its numbers.
Sources
- FuturesAI Insights: BTC Funding Rate Analysis (dated March 8, 2026; read November 8, 2026)
- FuturesAI Payback: partner exchange fee table (read November 8, 2026)
