Left Of Meta Research
Onchain forensics

The sharpshooter
that sells to itself

On July 11, 2026 a Hyperliquid wallet opened a $900K MINA long 15% below market and was up $161K within a minute — the kind of fill wallet trackers flag as elite. We pulled the fills, the explorer transactions, and the funding trail. It is not an elite trader. It is a five-wallet operation manufacturing one. July 2026. Descriptive analysis, not investment advice.

Executive summary

$900,000Single MINA fill · 15% below mark
67 msFrom resting sell to snipe
4 eventsSame counterparty wallet, 4 weeks
$3.55MWithdrawn by B against $500K in

One minute on a dead book

MINA-USDC on Hyperliquid trades ~$1.3M a day. For 52 minutes before the event the book printed 300–2,000 units per minute. Then, in the single minute of 06:53 UTC, 26.3M units traded across 114 fills, the price wicked from $0.0431 to $0.036824 — and closed the same minute back at $0.0430. The wick never appeared on any other venue, and Hyperliquid's oracle price never left ~$0.0435: this was one venue's order book emptying and refilling, not price discovery.

The natural reading — a leveraged long force-liquidated into a thin book — fails on the data. Hyperliquid flags liquidation fills in both parties' fill records. No fill in this event carries the flag. What the records show instead:

The event, reconstructed from both wallets' fills

06:53:55.015  A fires ~43 taker sells, sweeping the entire bid book $0.0431 → $0.0391
              (~$50K notional was enough — the book was that thin)
06:53:55.082  A rests an OPENING SHORT: 24,439,918 MINA at $0.036825   (maker)
06:53:55.082  B lifts the entire block with a GTC limit buy            (taker)
              — same trade id 614021723573662, same size, same price, same millisecond
              — B's size = floor($900,000 / 0.036825): sized to a round dollar budget
06:54         price back at ~$0.0436; B is +$161,475 unrealized (+53.8% ROE)

Sources: userFills for both wallets (Hyperliquid info API); order payloads and signer addresses via explorer txDetails; tx 0xd89a19ea2cb97476da13043fae866401e70031cfc7bc93487c62c53cebbd4e61, block 1068402276. In the minute before the event the MINA book saw four small market-maker orders — an unrelated bot discovering, pricing and filling a 24.4M block in 67ms is not a market interaction.

Four wins, one counterparty

B's recent 2,000 fills (May 6 – July 11) sum to +$980,545 closed PnL on $4,282 of fees, with zero liquidations. Its four largest entries share one fingerprint: a flash-crash wick low on a dead book, printed at 10–16% below the price a candle earlier, on volume hundreds to tens of thousands of times the local median.

Date (UTC)MarketEntry notionalVs prior candleCandle volume ÷ medianOutcome
Jun 15XAI$864,977−9.9%×333+$110,810 in 25 min
Jun 17AXS$1,237,724counterparty: wallet A
Jun 27GMT$863,690−11.1%×2,841+$120,776 in 17 min
Jul 5POLYX$853,431−13.3%×10,043+$210,947 in 2 days
Jul 11MINA$900,000−15.6%×~26,000+$161,475 open

Attribution is mechanical, not statistical. Where B was the maker, Hyperliquid's fill record carries the taker's transaction hash; resolving those through the explorer names the aggressor who crashed the price into B's resting bids. AXS, GMT, POLYX: all three resolve to wallet A. MINA (where the roles were reversed and B took A's resting sell) matches by shared trade id. Four events, four weeks, three mechanisms of matching — one counterparty.

B — the "winner" · 0x8bae…3ab6dA — the loser · 0xd8d5…7349
Closed PnL, recent 2,000 fills+$980,545−$664,087
Liquidation-flagged fills025
Account value (Jul 11)$368,530$390,651
Markets tradednear-identical: ZEC · FET · AXS · XAI · GMT · POLYX · SAGA · spot @166 · tokenized-equity perps

Follow the money

Hyperliquid's ledger endpoint exposes every deposit, withdrawal, and transfer; Arbitrum's USDC transfer logs fill in the off-exchange hops. Together they close the loop.

B: one deposit, seven figures out

2025-07-10 05:40 — cluster wallet F_B (0x9458…ec04) withdraws from Hyperliquid$499,999
2025-07-11 08:33 — F_B wires it to B on Arbitrum$499,999
2025-07-11 08:35 — B deposits to Hyperliquid (its only deposit, ever)$499,999
Nov 2025 – Mar 2026 — B withdraws in six transactions$3,551,994
First $700K exits, 2.5 min after withdrawal, to clean wallet D_B (0x3828…da9e)$699,999

$499,999 is $500,000 minus Hyperliquid's exact $1 withdrawal fee: the seed money was itself another account's Hyperliquid withdrawal. F_B's ledger shows ~15 withdrawals in the same $X99,999 pattern — roughly $10M cycled since early 2025.

A: refilled on a timer

2025-11-21 12:33:48 — treasury F_A (0x7b8e…86b4) → A on Arbitrum$2,000,000.00
2025-11-21 12:34:58 — A deposits to Hyperliquid (70 seconds later)$2,005,959
2026-03-05 12:45:43 — F_A → A on Arbitrum$2,000,093.96
2026-03-05 12:46:58 — A deposits to Hyperliquid (75 seconds later)$2,000,093.96
Plus five UETH transfers from 0x8dafbe…35c0≈ $5.6M

F_A is an early Hyperliquid account (active Feb 2024; PURR and HYPE genesis allocations) repurposed as an external treasury. A has churned >$11M of deposits through a year of continuous losses. On 2026-03-05, B moved funds internally at 12:43:46, F_A wired A at 12:45:43, and A deposited at 12:46:58 — three wallets, two chains, three minutes.

Verdict

A single automated operation running at least five wallets: a designated loser (A) crashes thin Hyperliquid books and hands ~$850K–1.2M blocks to a designated winner (B) at 10–16% discounts; treasury wallets (F_A, F_B) refill the loser and seeded the winner; the winner's manufactured profits — $3.55M withdrawn against a single $500K deposit — exit to a clean wallet (D_B). Not a liquidation, not a fat-finger, not luck: wash-trading between coordinated accounts.

The structure fits two purposes, which are not mutually exclusive: track-record manufacturing — to any wallet screener, B looks like a generational talent, and followers who mirror it become exit liquidity — and value transfer/laundering, converting one pocket's deposits into another pocket's "trading profits" at ~15% friction per hop. Separating the two would require the parties behind D_B's downstream flows.

What to take from it

Copy-trade screeners rank wallets by realized PnL, hit rate, and drawdown. Every one of those numbers can be manufactured for the cost of the friction between two of your own accounts — and on books thin enough, the friction is a rounding error against what a following builds. Before mirroring a "sharpshooter," check the one thing this operation cannot fake: who is on the other side of its best trades. On Hyperliquid that check is free — fills, trade ids, and taker transactions are public. If the same address keeps appearing across a wallet's signature wins, you are not looking at a trader. You are looking at a stage play with a ticket booth.

Method & sources

All data public. Hyperliquid info API (candleSnapshot, userFills, userFillsByTime, clearinghouseState, userNonFundingLedgerUpdates) · Hyperliquid explorer RPC (txDetails, blockDetails; ~850 blocks scanned around the MINA fill) · Arbitrum One JSON-RPC, USDC transfer logs with block timestamps by binary search. Counterparty attribution: maker fills carry the taker's transaction hash; definitive matching via shared trade id in both wallets' fill records. Key addresses — A 0xd8d5b0f7219aed171efb5ae2ac3b9941206f7349 · B 0x8bae3527e5a33fa0cf184f37bc112d071463ab6d · F_A 0x7b8e0066791a0e4bd16c5cf1b02c97661ae086b4 · F_B 0x9458f423df6569f053358ee9906613cd69c1ec04 · D_B 0x38280137274987f4998b4d9dacc4137b91eada9e · UETH feeder 0x8dafbe89302656a7df43c470e9ebcb4c540835c0.

Descriptive on-chain analysis of pseudonymous addresses, based on public exchange and blockchain data retrieved July 11, 2026. Figures are as reported by the respective APIs at that time. Nothing here identifies any natural person or entity, and nothing is investment advice.