Hyperliquid's 7-Day Leaderboard: Two Very Different Risk Books

Six accounts sit at the top of Hyperliquid's public 7-day leaderboard this week, and together they booked roughly $282.6 million in realized PnL. The headline number is less interesting than its shape: the money splits into two clearly different kinds of risk-taking, and the ROI column is where the split shows.

The full set, as reported on 2026-07-20:

The dollar leaders are barely taking directional risk

The two largest accounts dominate the PnL column — $177.0M and $50.8M, about 80% of the combined total — yet their weekly ROI sits at 1.3% and 1.1%. Against equity bases of $13.48 billion and $4.69 billion, those are the return profiles of flow businesses, not conviction trades. Books this size earning low-single-digit weekly returns look far more like market-making, basis capture, and funding harvesting than leveraged directional bets. That is worth internalizing when leaderboard screenshots circulate: the biggest green numbers on Hyperliquid this week are almost certainly payment for providing liquidity and warehousing inventory, not for calling direction on BTC or ETH.

There's a second implication in the equity figures themselves. Two accounts holding a combined $18.2 billion on a single perp venue says something about where institutional-scale flow infrastructure now lives. Capital of that size doesn't sit somewhere it can't hedge, exit, and rebalance continuously.

The efficiency tier is where directional risk actually lives

The more interesting rows are further down. 0xd475…1a91 turned $53.0 million of equity into $7.66 million of realized PnL — a 15.2% week. 0xb83d…6e36 printed 12.0% on $109.8 million, and 0x856c…910d made 8.4% on $71.8 million. 0x8d68…52dc bridges the tiers: $38.3 million of PnL at 6.8% ROI on a $599 million book — large enough to matter in the dollar rankings, efficient enough that real directional exposure is clearly being taken.

Double-digit weekly ROI does not come from passive flow capture. These accounts are running concentrated positions, meaningful leverage, or both, and this week it worked. The leaderboard structure is a reasonable proxy for where discretionary perp risk is being expressed right now: not in the mega-books, but in the $50M–$600M tier, where a trader can still move size without becoming the market.

What a realized-PnL leaderboard can and cannot tell you

Two caveats keep this data honest.

First, these are realized figures over a seven-day window. They tell you what these accounts closed, not what they hold, and not what a similar strategy returns over a quarter. A 15.2% week is an outcome, not a run rate.

Second, a leaderboard is survivorship by construction. For every account that realized 12% this week with concentrated exposure, an unknown number of accounts ran comparable risk and don't appear anywhere. Ranking by winners systematically hides the loss distribution of the strategy class that produced them.

The useful read, then, is structural rather than imitative. The venue's largest capital is behaving like infrastructure — thin, steady returns on enormous books. The aggressive risk-taking that most retail traders imagine when they see a leaderboard is concentrated a tier below, in accounts one to two orders of magnitude smaller, where this week's ROI ran between 6.8% and 15.2%. Whether that tier's performance persists is exactly the kind of question a single week of realized data cannot answer.

Track live positioning and perp-flow data yourself at app.formion.ai.