While computing peak exposure across 9,733 Hyperliquid addresses, two of them came back with the same number — to the dollar. Peak exposure is a continuous figure accumulated fill by fill over hundreds of trades. Two independent traders landing on the same value is about as likely as two strangers quoting the same bank balance down to the cent.
0xbd1c84113c6deb5044be0c2221f9e9403811ff0e
peak exposure
$24,395,119
0xec7f9f5b31b8f5da6ac6c0973bbd496e16f9df39
peak exposure $24,395,119
Not rounded. That is the figure both wallets report.
Both traded only BTC. Both opened on 2026-06-25 and have run the same position structure since — opening, closing and reopening on the same days, with trade volumes within about a percent of each other.
But the fill counts differ. That detail decides the whole question. If this were one account being double-counted, the number of fills would be identical. Matching timing and size with different execution means two real wallets, one hand, one set of instructions — and the exchange splitting each order slightly differently.
Widening the search to every wallet that opened a BTC position within a day of 2026-06-25 and traded more than $20M returns 128 wallets. Ranked by profit, the top three are:
| Wallet | P&L | Volume |
|---|---|---|
0x237152b0… | $7.8M | $66.3M |
0xbd1c8411… | $5.4M | $28.0M |
0xec7f9f5b… | $5.3M | $27.8M |
The largest of the three runs a bigger position, but the same rhythm: same open day, same close day, same reopen. Together they account for $18.5M. On any leaderboard, that is three different people.
We are not going to dress this up as a brilliant call. These wallets were long BTC through a period when BTC rose sharply. Making money on that is not the part that needs explaining.
The part worth looking at is that three wallets did the identical thing on the identical days — and that the same cohort contains wallets that lost millions over the same window, on the same coin, with the same order type.
Running the exact-collision test across the whole database: 9,733 addresses produce 7 pairs whose peak exposure matches another address to the dollar. Checking each pair's behaviour, 3 are clean mirrors:
| Coin | Window | Wallet A | Wallet B | P&L A | P&L B |
|---|---|---|---|---|---|
| BTC | 2026-06-25 → 2026-09-21 | 0xbd1c8411… | 0xec7f9f5b… | $8.0M | $7.9M |
| HYPE | 2026-04-04 → 2026-08-20 | 0x0134c6a8… | 0x7176cec3… | $690K | $688K |
| ETH | 2026-07-23 → 2026-09-22 | 0x5b7cb815… | 0xf1e4a107… | $230K | $230K |
All of them share one profile: a single coin, a single unbroken position, 100% taker execution, matching open and close dates, and P&L within about 1%. The remaining pairs trade entirely different coins — those are genuine coincidences.
If you use any smart-money leaderboard to make decisions, this has three direct consequences.
There is no "same entity" field on a blockchain, so we state only what the data supports: these wallets are indistinguishable in timing and size. For deciding whether to treat them as independent signals, that is enough.
Every number here resolves on chain. The full round history for each wallet is public and needs no login:
Each page links out to the same address on Hyperliquid's own explorer, where the raw fills are. Other leaderboards hand you a ranking. We hand you every fill that ranking was computed from.
A round is one address in one coin, from the moment the position opens from zero until it returns to zero. Adds and trims along the way stay inside the same round, so a round corresponds to one decision the trader actually made.
We do not use realized P&L — a trader who never cuts a loser shows a 100% win rate. Everything here is marked to market: open positions are valued at the current mark, unrealized moves included.
Market makers are excluded (500+ fills a day under 30% taker ratio: quoting, not taking a view). Liquidation backstops are excluded — their win rates come from the liquidation spread, not from trading.