preview · November 20261.3 million accounts. 79 trade well.preview
Our screen, run on every Hyperliquid account from the public record: 317 pass, 79 survive a second read, 38 trade at our pace. Who they are, what their equity looks like, and how they trade.
This is a preview. The numbers are a first reading and may move before the full study in November. Accounts appear by the start of their address only.
Of 1.3 million accounts, 317 pass the screen. 79 remain after a second read, and 38 of those hold a position for five minutes or more: the pace our data and our first study start from. The selection holds out of sample: chosen with what was known in April, two accounts in three were in profit in the six months after, against one in three of the others.
The equity of the 38
Cumulative profit and loss of each account, in dollars, from 25 May 2025 or its first trade; the thin line is zero. Under each curve: total, days plotted, and the median time a position is held. These are the accounts to look at more closely.
The other 41 of the 79 hold a position for under five minutes. Their edge is speed, and our timeframes start at five minutes: they are left out of this page.
How the 38 trade
Median time a position is held, for each of the 38.
Does the selection hold
The test that matters. The screen is applied again as of 19 April 2026, using only what was known that day, and the accounts it picks are followed for the next six months.
| group | accounts | in profit after | median PnL after |
|---|---|---|---|
| picked by the screen | 277 | 62% | +$617 |
| all the others | 23,789 | 32% | −$66 |
| picked, trading often | 71 | 79% | +$23,025 |
| picked, trading rarely | 54 | 41% | −$21 |
After: 20 April to 19 October 2026. Medians, not means: two of the picked accounts lost more than $5 million each.
The selection works on accounts that trade often and is a coin toss on those that trade rarely. A smooth curve is not a promise.
What this does not say
Nothing about which account to follow. It measures how many pass a screen and what became of them, and it picks the ones worth a closer look in the full study. An equity line that was smooth for a year can end in a day, and two of the picked accounts did. Only perpetuals are read.
Our curves were checked against the figures Hyperliquid itself publishes for the accounts: on 30 of them, 24 agree within 5% and none is off by more than half.