At 08:07 UTC today, Binance’s BTCUSDT perpetual marked at 81,929.0773 against an index of 81,941.1533.
The perpetual was cheaper than spot by 12 dollars — a premium of −0.014737%.
At that same moment the contract’s current funding reading was +0.010000%. Positive, and pinned at the baseline. The perpetual is at a discount, and longs are still the ones paying.
That is not an API glitch. It is what the formula says.
Put this period’s numbers back through the formula
Binance publishes the arithmetic:
funding rate = premium index + clamp(interest rate − premium index, −0.05%, +0.05%)BTCUSDT’s interest rate is 0.01% per eight hours. Substitute that −0.014737%:
clamp term = clamp(0.01% − (−0.014737%), −0.05%, +0.05%)
= clamp(0.024737%, …) = 0.024737% ← nowhere near the bound, passes through
rate = −0.014737% + 0.024737% = 0.010000%Exact to the last digit. SOLUSDT at the same instant behaves identically: mark 112.6000, index 112.6075, premium −0.006661%, which substitutes out to 0.010000% — precisely what the endpoint returned.
This is the dead zone, and today sits inside it
As long as the premium index stays between −0.04% and +0.06%, the clamp term cancels it exactly and the printed rate is identically 0.01%, regardless of what the premium actually is.
BTC’s −0.0147% and SOL’s −0.0067% today both sit comfortably in the middle of that band. Which means:
A 0.01% reading is not “longs are mildly warm.” It is another way of writing “the premium is inside the dead zone.” Information content: zero.
The readings that carry information are the ones below 0.01%, because those require the premium to break under −0.04% — a perpetual at a visible discount. BTCUSDT had 209 such periods out of 790 this year. The full count is in 790 periods, not one above 0.01%.
Four venues printing the same number
At the same instant, side by side:
| Venue | Current reading | Interval | Annualised |
|---|---|---|---|
| Binance | 0.010000% | 8 hours | 10.95% |
| Bybit | 0.010000% | 8 hours | 10.95% |
| OKX | 0.010000% | 8 hours | 10.95% |
| Hyperliquid | 0.001250% | 1 hour | 10.95% |
All four annualise to 10.95%, not a decimal apart. Hyperliquid’s 0.00125% is exactly 0.01% ÷ 8 — it settles hourly, so its reading is divided by construction.
That tidiness is not four venues agreeing. It is four venues whose premiums all happen to sit inside their respective dead zones, printing the baseline value itself. Reading those four numbers as “strong market consensus” gets it backwards: they say the market is expressing nothing at all right now.
One case that does not reconcile, stated plainly
In the same pull, ETHUSDT marked at 2,668.1171 against an index of 2,668.9093, an instantaneous premium of −0.029683%. Run that through the formula and you should get 0.010000%. The endpoint returned 0.009345%.
A gap of 0.000655 percentage points. The formula is not wrong: funding uses the time-weighted average premium index across the whole period, not the instantaneous value at the second you called the API. BTC and SOL reconciled exactly because their premium tracked the period average closely; ETH was choppier this period, so the snapshot does not substitute out cleanly.
We spell this out because back-solving a full period’s rate from one second of data only works when conditions are calm. It is a useful sanity check, not a reliable predictor.
What it means for a position
A balanced two-legged book cares about one thing: has funding turned negative. It has not — all four venues printed positive on BTC, leaving you on the receiving side.
But today’s numbers make a second point: a discount is not the same as negative funding. The perpetual traded 12 dollars below spot and shorts were still the side collecting. Anyone planning to go long the perpetual whenever they see a discount would have come up empty today — turning funding genuinely negative needs the discount to run deeper than −0.04%, more than twice as deep as today’s.
What to watch next
- Whether the premium breaks the −0.04% floor of the dead zone. That is the precondition for negative funding, and the only line worth watching.
- Contracts with a zero interest rate, like BNB. It read 0.010527% today against an instantaneous premium of +0.060646%; with a zero baseline its dead zone relocates to zero, so the whole judgement needs a different set of numbers. There is a batch of these, counted here.
- Do not encode 0.01% as a market signal. It is a constant, not an observation.
Funding drifting toward zero in a range and turning negative in a downtrend is something this approach genuinely runs into — the 2022 row on our track record, where ETHUSDT returned 0.787256%, came from exactly that, and we did not hide it. Positive today does not mean positive next period.
Sources: mark price, index price, interest rate and the current funding reading all come from Binance’s public USD-M endpoint fapi/v1/premiumIndex (BTCUSDT / ETHUSDT / SOLUSDT / BNBUSDT), read at 2026-09-21 08:07 UTC; premium is computed as (mark − index) ÷ index; the cross-venue comparison comes from Hyperliquid’s public endpoint info/predictedFundings (one response carrying Binance, Bybit and Hyperliquid quotes with their intervals) and OKX api/v5/public/funding-rate at the same instant; the per-period history and the count of 209 negative periods come from fapi/v1/fundingRate, 2026-01-01 00:00 to 2026-09-21 00:00 UTC, 790 periods; the clamp formula and its ±0.05% bounds are from Binance’s public documentation. Past performance does not indicate future results.