Study · re-run every night · updated
Do crowded shorts get squeezed? Altcoins' funding rates and the week after, tested
On Oct 14, of our 100 most traded altcoins with a funding rate, 1 had shorts paying 50% a year or more and 0 had longs paying 100% or more. After 2,788 coin-days of crowded shorts since 2019, the coin did −3.0% against the other coins over the next 7 days on average, ahead 35% of the time; after crowded longs, +3.1%. Past the bar: crowded shorts, the next 7 days, behind the other coins. Past patterns, not financial advice.
Crowded shorts
−3.0%
next 7 days vs the rest
Crowded longs
+3.1%
next 7 days vs the rest
Today
1 · 0
crowded shorts · longs
§ 01 · Today
Today
as of · funding on Binance (Hyperliquid on the days Binance has none), Nov 28, 2019 to Oct 14, 2026
On a perpetual future the funding rate is what one side pays the other to keep the price near the spot price: above zero the longs pay the shorts, under zero the shorts pay the longs. The saying: when shorts pile in so hard that they pay through the nose, a short squeeze follows - and when the longs are the ones paying, the coin is about to dump. Here every day's funding of our 100 most traded altcoins with one (stablecoins, wrapped coins and Bitcoin itself left out), read as a year's rate - the exchanges' default, 0.01% every 8 hours, is about 11.0% a year.
Oct 14, 2026: 1 crowded shorts, 17 shorts paying, 82 normal, 0 hot, 0 crowded longs.
§ 02 · After it
What came after it
as of · from the reading day's close, against the day's other coins
the next 7 daysthe next 30 daysunder each band: its coin-days
| Funding | Coin-days | Next 7 days | Next 30 days | Held 7 days, a year | Held 30 days, a year |
|---|---|---|---|---|---|
| Crowded shorts shorts paying 50% a year or more |
2,804 | −3.0% 35% ahead | −9.4% 25% ahead | −73% t −3.0 ✓ | −54% t −2.4 |
| Shorts paying shorts paying, under 50% a year |
11,666 | +0.5% 46% ahead | +0.3% 42% ahead | most coins | most coins |
| Normal about the default, 0 to 29% a year |
27,666 | −0.0% 40% ahead | +0.2% 38% ahead | most coins | most coins |
| Hot longs paying 30 to 99% a year |
656 | +4.7% 33% ahead | +23.0% 31% ahead | −27% t −0.2 | +396% t 1.8 |
| Crowded longs longs paying 100% a year or more |
187 | +3.1% 31% ahead | +20.4% 34% ahead | −58% t −0.3 | +77% t 0.3 |
Next 7 and 30 days: the coin's move from the reading day's close less the average move of that day's other coins over the same days, averaged over the band's coin-days, with how often the coin was ahead. Held, a year: the luck test's own number - every coin within the hold after a day in the band, held together in equal stakes, day by day, against the other coins, as a year's rate (✓ past the bar of 3.0, both halves of the days agreeing; the halves on hover). "Most coins": the band's coins were most of the day's coins on the average day, so "the other coins" would be the few odd ones out, not the market - no luck test there.
Crowded shorts: after 2,788 coin-days the coin did −3.0% against the rest over the next week (−4.2% in the middle case), ahead 35% of the time. Past the bar: the shorts were right.
§ 03 · How it is tested
How it is tested
as of · methodology
A luck test (t) says how far a result sits from zero, in units of its own noise: about 2 is rarely luck for one test, but many tests at once need more, which is what the first cut is for.
A coin-day's funding is the day's payments on Binance's perpetual added up and spread over its 24 hours, Hyperliquid's on a day Binance has none - one venue a day, never the two blended - and all of it is paid by 16:00 UTC, before the day's close the coin is bought at. The 100 coins are the most traded that day by dollar volume over the 30 days before. The bands were fixed before any number was read: 50% a year or more for the shorts, 100% or more for the longs, the default in the middle.
Coins in one band cluster in time - whole markets turn crowded together - so the luck test does not count coin-days as separate evidence: it holds every coin within the hold after a day in the band as one portfolio, day by day, against the day's other coins in equal stakes (each side built alike, so neither earns the lift equal stakes in swinging coins give), the days as its cases (a Newey-West error over ten days). A band gets the test only where its held coins were under half of the day's coins on the average day (read from the bands alone, before any move). With 6 tests, a result needs a luck test of 3.0 or more with the older and the newer half of the days on its side. One change after the first read of our data, on Oct 14, 2026: the middle bands, held on most coins most days, had been tested against the few left over. Prices only, before trading costs and the funding itself.
§ 04 · Revisions1 dated
Revisions
as of · every change to its headline numbers, dated
This study is re-run every night on the whole history. It was first written on Oct 15, 2026; its numbers last moved on Oct 15, 2026. Each night its headline numbers moved is listed here, newest first.
- Against the other coins a week on, as a year: crowded shorts −73% · crowded longs −58% · past the bar: crowded shorts, the next 7 days, behind the rest
§ 05 · More
More research
as of