Funding Breadth Across 50 Alt Perps: 86 Percent Positive, 54 Percent Actually Market-Priced

Measured first-hand over 61 complete UTC days, 2026-08-10 to 2026-10-09, pulled 2026-10-10 16:41 UTC.

Funding breadth is the share of a basket of perpetuals paying positive funding on a given day. It is meant to answer a question a single Bitcoin funding number cannot: is leverage one-sided across the whole market, or only in the one contract everybody watches. I built the measure, and then the measure half destroyed itself, because most of what it was counting was not a market price at all.

What share of alt perpetuals actually pay a market-determined positive funding rate?

Over 61 complete days on Binance USD-margined perpetuals, a median of 86 percent of the 50 highest-volume alt perps paid positive funding on any given day. But 54.4 percent of the 13,046 alt funding settlements in that window printed a value exactly equal to that contract's own clamped interest rate, which means the premium index sat inside a 10 basis point band and no market premium was measured at all. Reclassify those and the median falls to 54 percent, with a 95 percent moving-block bootstrap interval of 48 to 56 percent. So the honest answer is a coin flip, not a one-sided book, and the 32 point difference between the two numbers is an artifact of the funding formula rather than a fact about positioning.

Why a positive funding print is often not a positive premium

Binance computes funding as F = [P + clamp(I - P, -0.05 percent, +0.05 percent)] / (8 / N), where P is the average premium index (time-weighted, or a plain arithmetic mean on contracts funding hourly), I is the contract interest rate (default 0.01 percent per 8 hours, 0.03 percent per day, but set to zero on some contracts), and N is the funding interval in hours. The algebra matters: when the premium index sits within or exactly at 0.05 percent of the interest rate, the clamp term returns I - P, the P cancels, and the funding rate collapses to I / (8 / N). The print is then the venue's own constant, carrying no information about whether longs or shorts were paying.

Athenum already measured how often that happens board-wide, 118,089 of 170,162 Binance settlements, 69.4 percent, printing the one constant value while the contracts kept trading. What follows asks a different question: what a rate of that kind does to a breadth statistic built on top of it. The 69.4 percent board-wide figure and the 72.8 percent control draw reported at the end of this piece corroborate each other; the 54.4 percent below is lower than both because this basket is the liquid top of the board.

That constant is interval-dependent, which is the trap. On an 8 hour contract it is 0.0100 percent. On a 4 hour contract the same formula divides by 2, so it is 0.0050 percent. My first pass tested only for 0.0100 percent and therefore missed the clamp on the 21 of 50 alt contracts that settle every 4 hours, which inflated the corrected breadth figure by 16 points before I caught it. Anyone reproducing this has to read the funding interval per symbol, not per exchange, which is a point Athenum measured across 2,536 contracts earlier this year. The interest rate is per symbol too: BNBUSDT runs an interest rate of exactly zero, so its clamped print is 0.0000 percent, and testing it against 0.0100 percent scores it as never clamped when it is in fact clamped on 41.2 percent of its settlements.

Method, in five steps

1. Pull every USD-margined perpetual with status TRADING from the public exchangeInfo endpoint and keep the USDT-quoted ones (525 of the 568 TRADING perpetuals on 2026-10-10), holding the quote currency fixed so a USDC-quoted contract cannot contaminate the comparison.
2. Rank by 24 hour quote volume, take the top 50 excluding BTCUSDT, and keep BTCUSDT as the reference.
3. Pull the full public funding history per symbol for the window. 13,075 alt settlements land on the 61 days where all 51 symbols report; 13,046 of them have a preceding print, which is what the interval test below needs.
4. Sum each symbol's funding per UTC day rather than averaging rates, because 29 of the 50 alts settle 3 times a day and 21 settle 6 times, so an average of rates silently weights the 4 hour contracts differently from the 8 hour ones. Two of them switched to hourly funding mid-window after a settlement hit the rate floor, so the interval is read per settlement from the gap to the previous print, not assumed per symbol.
5. Mark a settlement as clamped when it equals that contract's own interest rate times N divided by 8. Headline breadth counts any positive day; strict breadth counts a positive day only when at least one of its settlements was away from the clamp.

The two measures side by side

The gap between them runs a median 28 points and reached 72 points on 2026-08-23, a day when headline breadth read 98 percent and strict breadth read 26 percent. Across the window, 930 of the 2,521 positive symbol-days, 36.9 percent, consisted entirely of clamped prints.

Strict breadth as defined is the loose version of the correction: it keeps a symbol-day whose non-clamped prints were negative but whose clamped prints dragged the daily sum positive, and that describes 233 of the 1,591 strict-positive symbol-days, 14.6 percent. Requiring the market-priced component itself to be positive puts the median at 46 percent, above 50 percent on only 17 of the 61 days. The correction therefore has a range, 46 to 54 percent, and both ends are far from 86.

The measure only earns its keep after the correction

This is the part that decides whether funding breadth is worth tracking. Headline breadth agrees with the sign of Bitcoin's own daily funding on 58 of 61 days, 95 percent, so as a direction signal it is very nearly a redundant copy of the number you already had. Strict breadth agrees on 39 of 61 days, 64 percent. The corrected measure is the only one of the two that is ever telling you something a Bitcoin-only read does not.

Bitcoin is the exception, not the template

Ranked by the share of settlements sitting at the clamp, BTCUSDT is 48th out of the 51 contracts measured, at 17.0 percent, and ETHUSDT is 13.7 percent. The median alt sits at 48.6 percent, 12 alts are above 70 percent, and LUMIAUSDT printed the clamped rate on 99.5 percent of its 366 settlements. The contracts whose funding people actually read are the ones where funding is genuinely priced; the long tail is mostly quoting the exchange's interest rate back at you.

BNBUSDT is in that table as a warning: its interest rate is zero, so its clamped print is 0.0000 percent, which is also exactly what a contract with no premium at all prints. On a zero-interest contract the clamp is not cleanly separable from an inactive book.

What early October looked like, read correctly

Headline breadth fell from 84 percent on 2026-10-04 to 48 percent on 2026-10-09, its lowest reading in the window. Strict breadth over the same six days went 54, 42, 54, 56, 48, 44 percent, a far smaller move, and the gap on 2026-10-09 was 4 points against a median of 28. Two things happened at once, and the headline number cannot separate them, so here is the symbol-by-symbol transition.

The clamp did empty out: entirely clamped alts fell from 15 on 2026-10-04 to 3 on 2026-10-09. But of the 13 contracts that left the clamp, 7 went positive and only 6 went negative, so clamp exits explain just 6 of the 18 symbol fall in headline breadth. The larger part, 12 symbols, had been paying a genuine non-clamped positive premium and flipped negative. Alts paying negative funding went from 8 to 25. So this was a real move in positioning and not merely a measurement artifact, which is the opposite of what the gap alone suggests, and it is why strict breadth fell only 10 points while the headline fell 36: the clamp drained in both directions at once.

Bitcoin's own daily funding went -0.00183 percent on 2026-10-06 and -0.00325 percent on 2026-10-07 before recovering to +0.00821 percent on 2026-10-09, so for three days the two measures disagreed about direction. Note that 2026-10-09 is the minimum of the headline series but not of the strict series, whose low is 20 percent on 2026-08-24.

Limits worth stating

The basket is the top 50 by 24 hour quote volume as of the pull, applied backwards across the window, and every symbol had to report on all 61 days, so it is the liquid and continuously listed end of the board. That selection flatters the result. One random draw of 60 USDT perpetuals from the 474 outside that top 50, same window and same method, 59 of which reported on all 61 days, came back 72.8 percent clamped against 54.4 percent, with a headline median of 88.1 percent and a strict median of 30.5 percent. On the wider board, then, genuine positive breadth is a clear minority rather than a coin flip, and 54 percent is the optimistic end of this measurement rather than a market-wide average.

One venue, one quote currency, 61 days. A clamped print means the premium index was within 0.05 percent of the interest rate; it does not mean the contract was inactive. Two symbols, RLCUSDT and KAIAUSDT, changed funding interval inside the window, which is why each settlement's interval is re-inferred from the gap to the previous print rather than assumed per symbol. Assuming a fixed interval per symbol instead would have reported 53.9 percent clamped, with the 54 percent strict median unchanged; the 54.4 percent above is the per-settlement construction. Both series are strongly autocorrelated, ACF(1) of 0.64 on the strict series, which puts the effective sample nearer 14 than 61; the interval quoted above is a moving-block bootstrap for that reason, and an ordinary day-by-day resample would have reported a falsely tight 52 to 56 percent. The default interest rate is also not universal: Binance documents contracts such as ETHBTC with the interest rate set to 0 percent, which would put their clamped value at zero rather than 0.0100 percent.

Cross-exchange aggregation of funding, open interest and order flow across the venues we cover is what Athenum is built for, and its explainer on how perpetual funding is actually calculated walks through the same premium index and clamp terms used above. The 37 free calculators on the site need no account.

Jackson Ly for Athenum Analytics