The monitor aggregates current open positions for two screened public-address cohorts. Main perpetuals use 55 lower-turnover swing addresses; HIP-3 uses the union of that cohort and a 111-address S/A research pool. It reports long notional, short notional, address counts, net direction, and cohort unrealized PnL by market. The ranking is not protocol-wide open interest or a complete whale list: it describes only the monitored denominator shown in the tool and inherits survivorship bias from historical address selection.
What this tool answers
The page answers a narrow question: among the addresses we actually monitor, which main perpetual and HIP-3 markets carry the largest current long and short exposure? It preserves four different measurements instead of collapsing them into one score:
- Gross notional is long notional plus short notional.
- Net notional is long notional minus short notional.
- Address count shows how many monitored wallets hold each direction.
- Cohort unrealized PnL sums the current position-level values returned in the snapshot.
The separation matters. Ten small short addresses should not automatically outweigh one materially larger long position. In the reference research, money-weighted and address-count signals often pointed in different directions, so the table shows both rather than hiding the disagreement.
How the monitored cohorts are selected
The main-perpetual cohort starts from public leaderboard addresses and applies historical filters: account value of at least $30,000, monthly volume of at least $2 million, monthly PnL of at least $5,000, positive all-time PnL, turnover no higher than 80 times account value, and an observed median holding period of at least 24 hours. Fifty-five addresses passed the saved screen.
HIP-3 adds a separately maintained S/A research pool. Its scheduled scan queries each selected address across HIP-3 DEX namespaces, records request health, and then unions those positions with the swing cohort. Duplicate address-market positions are removed before aggregation.
These are research rules, not neutral sampling. They overrepresent historically successful and active accounts and can miss a new, private, low-history, or recently funded large trader.
How net direction is classified
The tool computes imbalance = (long notional − short notional) ÷ gross notional. A market is net long above +15%, net short below −15%, and mixed inside that band. The threshold prevents a small dollar difference from being shown as a decisive directional signal.
This classification describes exposure at the snapshot time. It does not reveal whether positions hedge spot inventory, options, another venue, or another address. It also does not prove that a wallet is informed.
Coverage boundary
Hyperliquid’s public account query begins with a user address. It does not return one complete endpoint containing every holder for a market. A defensible market-wide ranking therefore requires an address universe, scheduled per-address queries, failure tracking, storage, and repeated refreshes. This first release publishes the denominator we can reproduce instead of relabeling it as “all whales.”
Use the market links to inspect the actual ranked addresses behind any row. Use the CSV export for independent calculations, and keep the snapshot date with any quotation so the number does not lose its time boundary.