How to Track Hyperliquid Whales: What the Raw Leaderboard Really Shows

Most guides to Hyperliquid whale tracking list a few websites. This one starts from the data those websites read. Hyperliquid publishes a leaderboard file and a public API that needs no key, so anyone can check what a tracker is really showing. We downloaded the raw file, sorted every large account using the rules behind our Hyperliquid whale board, then read those accounts one by one to see what they held.

Everything below comes from that work unless a source is linked. The leaderboard was read at 12:40 UTC on November 5, 2026, and the numbers move every hour. Our position tracking began on November 4, 2026, so it can describe a snapshot and little else. We make no claim here about trends or win rates. For background on the chain itself and bridge flows, see our Hyperliquid and Ethereum bridge data. For famous trades and the wider derivatives market, see how whales use options and perpetuals.

What does the raw Hyperliquid leaderboard actually contain?

The file lives at stats-data.hyperliquid.xyz/Mainnet/leaderboard. Each row holds an address, an optional display name, an account value, and profit, return and volume for four windows: day, week, month and all time. At our read it held 48,471 rows with a combined account value of $48.18 billion.

That is far from everyone. HyperTracker says it tracks more than 1.6 million wallets, which would make the leaderboard about 3% of them. We could not find a published entry rule, so we tested the file. Every one of the 48,471 rows had either $100,000 or more in account value, or $10 million or more in all-time volume. No row broke that rule. The lowest values among low-volume rows sat at $100,000.00 and the lowest volumes among low-value rows sat just above $10 million, which is a clean edge. Treat this as our measurement, not Hyperliquid's statement.

This matters later. A small account that lost its money and never reached $10 million in volume is simply absent. The board starts filtered before any tracker touches it.

How many $1M+ accounts are real position-taking whales?

Fewer than half, and they hold a small part of the money. We applied the same rule our collector uses, in order: account value of $1 million or more; all-time volume above zero; monthly volume no more than 100 times account value. Accounts with volume but no trades this month are "quiet".

The picture is much sharper at the top. Sort by account value and 9 of the top 10 accounts have zero volume, and the tenth has not traded this month. In the top 25, 17 never traded, 3 are quiet, 3 are market makers and just 2 are active whales. In the top 100, 67 never traded.

What is in the 934 accounts that never traded?

The board's volume field reads as combined volume, so zero means no perp or spot volume on record. Our rule calls these "system, vault and spot-only" accounts, which was a reasoned guess. So we read every one of them, then checked the staking and vault deposits of the 172 with the most value still unexplained.

The largest account of all shows a value of $15.29 billion with no trading history. It holds 499.66 million units of one spot token, USOL, against a total supply of 500 million in Hyperliquid's own spot data, or 99.93%. Valued at mark prices, all of that account's spot balances came to about $52.8 billion when we read them, most of it USOL, not $15.29 billion, so the board's figure does not even equal units times price for this kind of holding. This is a ledger reading of token reserves. Nothing about it describes a trader's capital.

Three accounts in this group hold $23.39 billion, more than half of all value at the $1 million bar. Any tracker that sorts by raw account value and calls the top of the list "whales" is mostly displaying token reserves and staking balances.

Why does this change what a tracker shows?

Filtering changes three things a reader sees.

The ranking. By account value, the list is reserves and stakers. By position size it is a different group entirely: at 12:44 UTC, the top 10 of our 727 wallets held 30.0% of whale position dollars, and the top 100 held 72.3%.

The long and short picture. Market makers carried $298.57 billion of the $359.60 billion month volume at the $1 million bar, or 83%, against $61.03 billion for 1,187 active whales. They typically quote both sides, so their inventory at any moment says little about a view. Leaving them in a whale long and short count mixes inventory with positioning. Our 727 wallets held $3.14 billion long and $3.47 billion short at 12:44 UTC. Hyperliquid's main-market open interest was $10.85 billion nine minutes later, so these wallets were around 29% of longs and 32% of shorts.

The profit table. Accounts with no volume can still rank high. Among the top 100 accounts by all-time profit at the $1 million bar, 44 had never traded. 699 of the 934 never-traded accounts showed more than $100,000 of all-time profit or loss, up to $411.7 million. Their "profit" is value change, as the next sections explain.

You can rerun the classification. This is the code we used for a later check, which returned 1,180 active, 928 never traded, 557 quiet and 87 market maker accounts from 48,454 rows at 13:26 UTC, a normal drift within an hour.

import json, urllib.request

URL = "https://stats-data.hyperliquid.xyz/Mainnet/leaderboard"
req = urllib.request.Request(URL, headers={"User-Agent": "your-app-name"})
rows = json.load(urllib.request.urlopen(req, timeout=180))["leaderboardRows"]

buckets = {}
for r in rows:
    value = float(r["accountValue"])
    if value < 1_000_000:
        continue
    w = dict(r["windowPerformances"])
    all_time, month = float(w["allTime"]["vlm"]), float(w["month"]["vlm"])
    if all_time <= 0:
        kind = "never traded"
    elif month / value > 100:
        kind = "market maker"
    elif month > 0:
        kind = "active"
    else:
        kind = "quiet"
    b = buckets.setdefault(kind, [0, 0.0])
    b[0] += 1
    b[1] += value

print(len(rows), "rows")
for kind, (n, usd) in sorted(buckets.items(), key=lambda x: -x[1][0]):
    print(f"{kind:14s} {n:6,d} accounts  ${usd / 1e9:7.2f}B")

How do you read a Hyperliquid wallet's public record, step by step?

All of these are POST requests to https://api.hyperliquid.xyz/info with a JSON body and no key, as set out in the info endpoint docs and the perpetuals section. Swap in the address you are studying. Order matters, because the early answers change how you should read the later ones.

curl -s https://api.hyperliquid.xyz/info \
  -H 'Content-Type: application/json' \
  -d '{"type":"clearinghouseState","user":"0xADDRESS"}'

Step 1: Find out what the address is

Send userRole. It returns user, agent, vault, subAccount or missing, and a sub-account answer names its master. Read a vault as a pooled strategy and a sub-account as a slice of something bigger. This call is expensive, weighted at 60 against a budget of 1,200 per minute, so use it for the addresses you care about instead of everything.

Step 2: Check the account mode

Send userAbstraction. This is the step almost no guide mentions, and it changes how you read everything else. Hyperliquid's docs say that under unified account and portfolio margin, balances should be read from the spot state, and that "individual perp dex user states are not meaningful." In our random sample of 200 active whales, 141 were on unified account and 44 on portfolio margin, 92.5% between them. The other 15 returned other modes.

Step 3: Read main-market positions

Send clearinghouseState. Each item in assetPositions gives the coin, signed size in szi (positive long, negative short), entryPx, positionValue, unrealizedPnl, marginUsed, leverage, liquidationPx and cumFunding. Keep the position fields and be careful with marginSummary.accountValue, for the reason in Step 2.

Step 4: Read the other markets

Send perpDexs for the list, then repeat clearinghouseState with a dex value for each. A default read covers only the first perp dex. Builder-deployed markets show as prefixed coins, such as xyz:XYZ100 in the docs. We show below how much this leaves out.

Step 5: Read what else the account holds

Send spotClearinghouseState for spot tokens, delegatorSummary for staked HYPE and userVaultEquities for vault deposits. An account with $20 million of value may hold $19 million of a token and run a small perp book, or the reverse. Those are different accounts with the same headline number.

Step 6: Read the fills

Send userFillsByTime with a startTime. The useful fields are dir (such as Open Long or Close Short), startPosition, closedPnl, fee and crossed, which separates fills that took liquidity from resting orders that were filled. The limits are firm: 2,000 per response, and only the 10,000 most recent fills are available. A busy account's history runs out quickly. Set aggregateByTime to true to merge partial fills.

Step 7: Separate profit from money moving

Send portfolio for account value and profit history, userNonFundingLedgerUpdates for deposits, transfers and withdrawals, and userFunding for funding paid or received. A jump in value on the same minute as a deposit is not a win.

Step 8: Look for related accounts

Send subAccounts on the master address. It returns each sub-account with its positions and spot state in one response, which is the fastest way to see a whole book, and Trap 4 below shows why this step is needed. For the method of linking addresses without a stated relationship, and why it often fails, see our guide on wallet clustering and the lesson on who owns a wallet.

The table explains how trackers are built. Live WebSocket feeds stop at 10 users per IP, so following hundreds of wallets from one machine means polling. Position reads are cheap and fill reads are not. Our own collector reads about 7 wallets a second, or 840 weight a minute, which gets through the whale list in about three minutes and repeats every five. That design has a cost, covered next.

What hidden traps make a whale wallet misleading?

Trap 1: The headline balance is not trading capital

The whale bar uses the board's account value. That number can include things no one is trading with. The docs say account value includes unrealized profit and vault balances, and our reads show staked HYPE and spot tokens in it too. Across all 1,187 active whales, the board showed $6.24 billion. The perp summary showed $1.18 billion, and spot balances at mark came to $3.78 billion. Those last two overlap under unified accounts, so they cannot simply be added. The honest statement is that no single public field equals "money this account trades with".

Trap 2: Holding nothing right now

526 of the 1,187 active whales held no main-market position at all when read, and 490 of the 561 quiet ones. A whale board shows who is exposed at this moment, and many large accounts are flat.

Trap 3: Markets a default read does not cover

Of our 200 sampled whales, 70 held positions on builder-deployed markets, 35% of the sample, across the 10 additional perp dexes listed at the time. Those positions came to $321.4 million against $746.9 million on the main market, about 30% of those two figures combined. The fills agree: of $251.3 million in sampled perp fills over 24 hours, $72.9 million, or 29.0%, was on those markets. Spot added another $14.3 million. Our own board covers main-market perpetuals only, so this blind spot is ours too.

Trap 4: Sub-accounts are missing from the board

Hyperliquid's docs allow up to 10 sub-accounts after $100,000 in volume, then one more per additional $100 million, up to 80. We expected to find them on the leaderboard as separate rows. We found none. Of 275 addresses we ran through userRole, including the top 50 by value and 150 active whales, not one came back as a sub-account. So we went the other way and asked our 200 sampled whales for their sub-accounts. 61 of them had some, 234 sub-accounts in total, and none of the 234 appears anywhere in the 48,471-row file.

This matters because those accounts hold positions. 20 of the 61 had positions in sub-accounts when we read them, 216 positions worth $117.8 million against $746.9 million on the same 200 main accounts. Any tracker that starts from the leaderboard and reads only those addresses, ours included, cannot see them. A long on a main account may have a short beside it in a sub-account, and the reader sees only the long. The subAccounts request is the direct public way to check. We did not test whether a master's board value or profit includes its sub-accounts.

Trap 5: Hedges elsewhere

This one has no fix from public data. A large short here may be the other side of spot held in cold storage, a long on a centralized exchange or an options structure. Nothing on Hyperliquid can show that. Inside Hyperliquid we can at least see part of it: at 12:44 UTC, 157 of our 727 wallets held longs and shorts at once, $2.60 billion of gross positions that net to $1.97 billion. Reading any one of those positions alone would miss the point of it, and 308 wallets held a single coin, which may or may not be the whole story.

Trap 6: Vaults among the whales

Vault leaders must keep at least 5% of the vault and earn a 10% profit share, so a vault's positions are a managed strategy with depositors' money. HLP, the protocol vault, runs market making and performs liquidations, so its book is not a directional view. The public vault file lists 9,478 vaults. Only 23 sit at $1 million or more on the board, and a volume filter catches few of them: 16 pass as active whales, 4 as market makers, 2 as quiet and 1 as never traded. That includes an HLP child account. The word "vault" in our method note covers only the last case. Their effect on our board is small, with 15 vault addresses holding $97.8 million, or 1.48% of position dollars, and yet a reader of a single wallet page needs to know. Vault deposits raise a depositor's account value too, without any position showing on that address.

Trap 7: Moves that close between reads

A tracker that polls every five minutes sees snapshots. A position opened and closed between two reads never appears. To size this, we reconstructed positions from 24 hours of fills for the same 200 wallets. Of 200, 84 traded. We found 47 position rounds that started from flat and reached $100,000 or more, and 25 of those closed inside the window.

So most large rounds were visible, and a real minority were not. The bigger distortion is in the event log. In our first 17.4 hours of events, 1,263 of 1,459 logged moves were increases or reductions, 86.6% of the count. Only 196 were opens, closes or flips. In that window most logged moves were resizes, and "opened" alerts described a small slice of them.

Trap 8: Unrealized and realized figures read as one number

Hyperliquid's docs define PnL as account value plus net deposits, and say the graphs sample every 15 minutes and on deposits and withdrawals, and that they are not meant for precise accounting. A direct test: of the 116 sampled whales with no fills in 24 hours, 49 held positions, and all 49 still showed a non-zero day profit, with a median size of $56,206. Open positions moved and funding accrued, so the number changed with no trade. We also compared the board with the portfolio request for 50 accounts. Allowing 5% tolerance, the board's all-time profit matched the combined allTime figure in 33 cases and the perp-only perpAllTime figure in 14. The two differed by more than $100,000 in 35 of the 50, by up to $13.9 million. Volume behaved the same way, matching the combined figure in 44 cases and perp-only in 16. So board profit and volume are not purely perpetual trading, and perpAllTime is the closer read when that is what you want.

Three practical consequences follow:

• A large profit on the board can vanish without a single trade. Across our 727 wallets at 12:44 UTC, 2,015 positions were up $407.9 million and 1,756 were down $342.1 million, all of it unrealized, for a net $65.8 million.
• For realized figures, add closedPnl from fills, subtract fees, and include funding. Our sample's fills showed $0.81 million of closed profit against $0.07 million in fees for that one day.
• Field values can be odd. 141 of 1,273 accounts with month volume read a month profit of exactly zero, some with a non-zero week figure. We report these as read, and we would not rank by month alone.

Trap 9: Liquidation prices that move

The docs state that a cross-margin position's liquidation price can shift with funding and with unrealized profit in other positions. On our board at 12:44 UTC, 3,581 of 3,771 positions were cross margin, $6.18 billion of $6.61 billion. So most liquidation prices are moving estimates that depend on the rest of the account, and 1,051 positions, 27.9%, had no liquidation price at all when read. A "liquidation level" on a chart is a reading at one moment.

Why does the top of the leaderboard reflect survivorship?

Survivorship bias is what happens when failures are removed from the sample before you measure it. Brown and co-authors showed in 1992 that a sample cut by survivorship can create the appearance of predictable performance in mutual funds, because high-volatility managers who survive look skilled. Hyperliquid's board has the same structure, and a current-balance bar makes it stronger.

This table does not show that big accounts trade better. It shows that accounts which made money tend to have big balances, which is close to a tautology. Sorting by today's balance selects the winners after the fact. Four more readings from the same file make the point:

• The missing losers. 917 traded accounts show an all-time loss of $1 million or more. Only 119 of them still hold $1 million, so 87% would never appear on a $1 million whale list. The largest all-time loss on the file is $128.9 million, on an account now holding less than a cent.
• The missing winners too. 2,234 traded accounts show $1 million or more of profit, and 935 still hold $1 million. A balance bar also drops people who took profit and withdrew, so it does not simply flatter the record. It distorts in both directions.
• Long record against recent record. Of the 1,834 traded accounts at $1 million or more, 83.5% are positive all time. Of the 1,273 with month volume, only 581, or 45.6%, were positive over the month. A long record and a recent record can look very different.
• The sum does not add up. Perpetuals are close to zero-sum before fees, and yet traded accounts on this file sum to $16.19 billion of gains against $5.47 billion of losses. Part of the gap is losers absent from the board, and part is non-trading value change such as spot and staked holdings. We cannot split the two from this file.

Did this month's leaders keep leading?

We ran one small test. The file gives both week and month windows, so we took month profit minus week profit as an earlier period, ranked the 1,273 accounts on it, and looked at the latest week.

The top 100 from the earlier period made $763.4 million then and $105.6 million in the latest week, and 41 of them were also in the latest week's top 100. Big accounts produce big dollar numbers in any direction, so overlap in a dollar ranking is partly a size effect. A proper test needs many non-overlapping periods and a record of accounts that left. Our tracking is days old, so we cannot run it yet, and we will not imply otherwise. The lesson on measuring results honestly sets out what such a test requires.

Why is copying Hyperliquid whale positions hard?

"Hyperliquid copy trading" is a common search, and several services offer it. This section describes the mechanics, with no view on whether anyone should do it. We looked at the same question on Ethereum spot in our whale copy simulation. Perpetuals bring a different set of frictions.

You see the position, not the reason

The record shows size, side and entry. It leaves out whether the position is a hedge, part of a basis trade, a loan-funded structure, a test of liquidity or a mistake. Because 157 wallets run both sides at once, one leg copied alone is a different exposure from the book it belongs to.

Exits matter as much as entries, and size changes them

Positions that a whale holds can be hard to leave at a reasonable price, and a copier's smaller size faces different conditions from a whale's. Whale positions in the largest coins were large relative to the book: at our reading, ETH showed $1.79 billion across positions of $100,000 or more against $55.7 million of depth within 1% on the thinner side, BTC $1.68 billion against $124.1 million, and HYPE $1.12 billion against $11.2 million. Across the 130 coins with a position of $100,000 or more, the median depth within 1% was about $102,000. Smaller coins have a far thinner book than the ones most people watch.

Leverage does not copy across

At 12:42 UTC on November 5, among positions of $100,000 or more, the median leverage setting was 10x and the 90th percentile was 20x. But with cross margin, the setting is not the real risk. The docs state that a cross position's liquidation price does not depend on the leverage set, because it draws on the whole account. A whale at "10x" with idle collateral and a copier at 10x with none hold very different risks.

Costs and behavior

Fees scale with volume. Hyperliquid's schedule starts at 0.045% taker and 0.015% maker and falls to 0.024% and 0.000% at the top volume tier, with separate maker rebates. A copy order that crosses the spread from a low tier pays a different cost from a large account's resting order, and funding accrues on each holder's own timing. Separate from cost, there is evidence about behavior. In a Management Science experiment, Apesteguia, Oechssler and Weidenholzer found that showing people others' success raised their risk taking, and that an option to copy directly raised it further. That was a lab setting, not Hyperliquid, and the authors' conclusion of excessive risk taking comes from that design.

Hyperliquid's own vaults are a different arrangement from mirroring a wallet. Depositors share in a leader's results and the leader keeps at least 5% in the same vault. The docs also note that legacy vaults lack spot and builder-market trading, and vaults carry their own risks and rules. We describe the structure only.

How do Hyperliquid whale trackers compare?

We opened each product's own page on November 5, 2026 and report only what we could confirm. Some are built for different jobs, and several go further than we do. Features and prices change, so check before relying on this.

A reasonable way to choose: if you want a read on every wallet and a developer API, a broad indexer fits. If you need many exchanges, a derivatives data platform fits. If you need a raw transaction record, use an explorer. If you want to study one account, the eight steps above work with no product at all. Whatever you pick, run the zero-volume check from earlier against its top accounts.

What can't our Hyperliquid data show?

Holding competitors to this standard means stating ours plainly. As of this writing:

• Leaderboard accounts only. Our list comes from Hyperliquid's public leaderboard file. An account that is not on that file is not on our list, however large it is.
• Main-market perpetuals only. We do not read builder-deployed markets or spot positions, so about 27% of the sampled position dollars sit outside our board.
• No sub-accounts. We start from the leaderboard, which leaves them out, so their positions are not on our board. They added about 10% to the position dollars in our sample.
• Vaults can pass the filter. Active vaults at $1 million or more can appear as whales. 16 did at this read.
• The balance is imperfect. The $1 million bar uses the board's value, which includes holdings beyond trading capital. The Account figure on our wallet pages comes from the perp summary, which understates accounts on unified margin.
• Five-minute reads. Short rounds can be missed, and a close's profit is the last unrealized figure we read, which is why we label it "about".
• Days of history. Tracking began on November 4, 2026. There are no trends, win rates or backtests here yet.
• No identities. We do not name the people behind addresses. A display name is whatever the account chose, and 85 of the 2,768 accounts have one. 35 of these whale addresses also appear in our separate Ethereum whale list. That shows the same address on two chains and no more.

Frequently asked questions

Is there an official Hyperliquid whale tracker?

Hyperliquid publishes a leaderboard and a public API, and every tool above builds on data like that. We found no published definition of "whale" from Hyperliquid, so every tracker's list reflects its own rules. Ask to see them.

How often does the leaderboard update?

We read the file hourly and see it change between reads. Hyperliquid does not state a schedule in the docs we read. For current positions, use clearinghouseState instead of the file.

Can you see a Hyperliquid wallet's full trade history?

Only the most recent 10,000 fills are available through the API, per the docs, at 2,000 per response. For very active accounts this can cover a short period. Older records need a tool that stored them earlier.

Does a large position mean a whale is confident?

Not on its own. It can be a hedge, a leg of a larger structure, a vault's strategy or market maker inventory. Size is what the record shows. The reason behind it is not recorded anywhere public.

Are market makers bad data?

No. They are a real and large part of the market, with 83% of month volume at the $1 million bar. They answer a different question, so a whale positioning count is clearer without them. Our 100× line is a judgment call, and some directional traders with high turnover are caught by it.

Bottom line

The raw Hyperliquid leaderboard is an extraordinary public record, but it is not a list of whales. On November 5, 2026, one third of its $1 million accounts had never traded and they held 77.8% of the value. 86 market makers made up 83% of the month's volume. After removing both, 1,748 traders remained, 1,187 traded that month, and 727 wallets held a position worth reporting.

A clean list is the start. The balance, the profit figure, the position read and the leverage setting each mean less than they appear to, for reasons that can be measured. Reading a wallet well means checking its role and account mode, covering sub-accounts and the other markets, separating realized from unrealized, and accepting that hedges elsewhere stay hidden. The top of any ranking also reflects who survived. Public data can tell you what an address held at a moment in time. Why it was held, and how it ends, stays outside the record.

Related reading

Hyperliquid & Ethereum: Whale Bridge Flows

Capital moving between the two chains, Hyperliquid's architecture and the limits of cross-chain tracking.

How Crypto Whales Trade Options & Derivatives

Deribit, GMX and Hyperliquid, famous trades, liquidation cascades and regulation.

Copying Whale Trades: A Simulation

Latency, slippage and whale types in Ethereum spot, the companion to the frictions measured here.

Which Ethereum Whale Wallets Are Profitable?

A 369-wallet profit study on Ethereum, and a useful contrast with leaderboard figures.

On-Chain Forensics & Wallet Clustering

Methods for linking addresses and the patterns that defeat them.

How to Read a Whale Wallet

The same reading habits applied to Ethereum wallets.

All whale derivatives → Hyperliquid whale ranking → BTC whale positions → Lesson: following a wallet → Lesson: spotting noise →