Methodology

Who counts as a whale

• A whale is an address with open perpetual positions worth $10M or more in total: longs and shorts across all coins, valued at the current mark price.
• To avoid flickering around the threshold, an address stays a whale until its positions drop below $5M.
• Addresses from $2.5M are watched: they are not counted in whale stats, but we notice the moment they become whales.
• Only crypto perpetuals on the main Hyperliquid exchange are covered. Stocks and commodities (HIP-3 markets) are not included yet.

Right now: 168 whales and 344 watched addresses.

How we find and track them

• Once a day we check every account from the Hyperliquid leaderboard with a balance of $250K+.
• In real time we read the stream of all trades on the exchange. Every trade includes the addresses of both sides, so any order of $100K+ puts its address on our radar within seconds.
• When a whale trades, we re-read its positions within seconds (bots that trade non-stop — at most every 30 seconds). Otherwise positions are refreshed every 10 minutes.
• Each snapshot is compared with the previous one, and the difference becomes a whale move: opened, added to, reduced, closed or flipped a position. Moves under $50K are ignored.

Directional vs all whales

Not every large position is a bet on price. A fund that holds $500M of BTC elsewhere and shorts $500M on Hyperliquid has zero net exposure — yet on Hyperliquid it looks like a giant whale short. A market maker’s short only means clients were buying from it. Counting such positions as “whale sentiment” distorts the picture, so the site has two views:

• Directional (default) — only whales whose positions look like a bet on the price going up or down. It answers the question: what do large traders who actually bet on direction think?
• All whales — everyone with $10M+ of positions, including hedges, baskets, market makers and vaults. It shows the full picture of risk: a hedge short can still be liquidated, and a squeeze of those shorts means forced buying.

Right now directional whales are 65% long, while all whales combined are 41% long. Current whale positions by type:

Switch between the views on the overview, coin and whales pages.

Whale types

Each whale gets exactly one type. The rules are checked in this order:

1. Vault — the address is a Hyperliquid vault, including HLP and its sub-vaults. Vaults are algorithms that usually take the other side of the crowd.
2. MM / HFT — monthly volume of at least 150× the balance with 15+ open positions at once; or, after at least 6 hours of observation, 70%+ of volume done as a maker with daily turnover of 5× the balance or more. Their positions are inventory left over from serving other traders, not a view on price.
3. Hedged — two or more positions where longs and shorts differ by less than 20% of the total.
4. Basket — five or more positions on the dominant side, each at least 3% of the total, with the book skewed at least 60% to one side. Typical of portfolio hedges and basis or funding trades: shorting “the whole market” against assets held elsewhere.
5. Directional — everything else: a clear bet on price.

These are heuristics. A trader shorting six coins with 20× leverage may be a genuine bear rather than a hedger, yet end up in “Basket”. That’s why every whale carries a visible type label and you can always switch to “All whales”.

What the numbers mean

Long / short

Share of long notional in whale positions, valued at mark price.

Net change, 24h: bought / sold

How much whales bought minus how much they sold over 24 hours — for the whole market and for each coin. Buying means opening or adding to a long, or closing a short; selling means opening or adding to a short, or closing a long. Price moves are not included — only the whales’ own trades, built from whale moves of $50K+.

Share of OI

Whale positions divided by twice the open interest: open interest counts one side of the market, while whales sit on both.

Funding

The market’s current hourly funding rate.

Leverage

Two numbers: the leverage set for the position and the effective leverage of the whole account (all positions divided by the balance). A 40× setting often hides only 5× of real leverage.

To liquidation

How far the price has to move to reach the position’s liquidation price. Under 10% is highlighted.

Large market orders

Single aggressive orders of $1M+ that hit the order book; all fills of one order are merged. An order split into several smaller ones won’t show up here, but it will appear in whale moves. Each order is compared with the coin’s trading volume on Hyperliquid over the previous 24 hours: from 1.5% of it the order is marked as an anomaly, from 3% — as extreme. The same dollar amount means very different things for BTC and for a small coin, so the share of daily volume is used rather than a fixed size. Orders that follow each other in the feed from one address in the same coin and direction, with pauses of up to 15 minutes, are highlighted as a series of one sliced order with a summary row; a series is checked for anomalies by its total. All orders over the last 24 hours are on the large orders page.

Squeeze risk and liquidation map

A squeeze is a chain of liquidations: the price reaches a cluster of liquidation prices, forced market orders push it further and trigger the next cluster. On Hyperliquid every position’s liquidation price is known exactly, so we can see where this fuel sits.

• Fuel — positions of tracked addresses (from $2.5M, any type: hedges get liquidated too) that would be liquidated if the price moved 2%, 5% or 10%. ↓ — longs, if the price falls; ↑ — shorts, if it rises.
• Book — order book liquidity in the same range: bids for a fall, asks for a rise.
• Fragility = fuel / book. From 1 the risk is High — liquidations exceed what the book can absorb; from 0.3 — Medium; otherwise Low. Levels above Low require at least $1M of fuel. The worst of the three ranges is shown.
• The liquidation map on each coin page shows the clusters: liquidations per 0.5% price step on top of order book liquidity. Where a colored bar is taller than the grey one, the book at that level can’t absorb the liquidations.
• Liquidation clusters are drawn as lines on the price chart: neighbouring levels with large liquidations are merged into one, clusters from $1M are shown, up to three per side. A solid line means the cluster is larger than the order book at those levels — a candidate for a cascade; dashed — the book can absorb it. These are current levels, not their history.

Limitations:

• Only addresses with $2.5M+ in positions are counted. Mid-size traders with leverage are not — the real fuel is larger.
• Hyperliquid softens cascades: positions over $100K are liquidated 20% at a time with a 30-second pause, and once margin falls below 2/3 of maintenance the position goes to the HLP vault without touching the book.
• The visible book thins out in volatile moments as market makers pull their orders. Beyond a few percent from the price Hyperliquid only returns an aggregated book, so depth there is approximate.
• The metric shows where the fuel is and how fragile the market is — not when a squeeze will happen. The thresholds are initial and will be calibrated on history.

Follow the flow

The Follow the flow page ranks coins by how unanimous directional whales are: the share of their money on one side. Coins from 60% get into the ranking, from 90% are marked as a strong consensus. To filter out noise, a coin needs $1M+ of whale positions, at least 3 whales on the dominant side, and the largest of them must hold less than half of that side — otherwise it is one trader’s bet. Next to the consensus: net change over 24h, the PnL of whales on that side, the squeeze risk for that side and funding.

Win rate

• A trade is one position cycle in a coin, from opening to full close. A flip closes one trade and opens the next.
• The result is realized PnL minus fees. Funding is not included yet.
• History comes from each address’s fills: Hyperliquid returns up to the last 10,000, and from then on we record everything ourselves. Positions opened before that history starts are not counted — their entry is unknown.
• Win rate is shown from 10 closed trades. Next to it we show trade PnL and profit factor: a high win rate with one large loss can still be a losing strategy.

Limitations

• We only see Hyperliquid. Hedges on other exchanges and spot holdings elsewhere are invisible.
• One trader can split positions across several addresses and sub-accounts; each address is treated separately.
• Positions refresh with a delay from a few seconds to 10 minutes; pages update every 30 seconds.
• HyperScan is an analytics tool. Nothing on the site is investment advice.