How Flow is made

• Every trade, live, from Hyperliquid's public trade stream: price, size, both wallets, and which side crossed the spread. Two identical connections run side by side so one can reconnect while the other carries on; duplicates are dropped by trade id. Rolled up per wallet, market and hour, and kept whole.
• Hyperliquid's leaderboard, hourly: account value, volume and PnL of about 48,000 wallets.
• Positions of the biggest wallets: the largest accounts, the biggest traders on the leaderboard and the biggest traders in our own data.
• Venue volumes every 5 minutes: each Hyperliquid market group, and the same coins' perpetual futures on Binance, Bybit and OKX from their public tickers.
• Labels: Hyperliquid's vault list, its HIP-3 deployers, Hypurrscan's community aliases, leaderboard names, and our hand-reviewed list of firm attributions.

Hyperliquid keeps only each wallet's last 10,000 fills, so history cannot be fetched later. We have collected every trade since —, and the longer windows fill in as time passes.

Once an hour is over, we fetch Hyperliquid's own candle for every market and compare its trade count and volume with ours. Completeness is measured, not assumed, and each Flow page shows it for what you are looking at.

The live figures load in a moment.

Every trade has a buyer and a seller. The maker had an order resting on the book; the taker crossed the spread to hit it. Hyperliquid marks which side crossed, and we checked that against both wallets' own fills.

A wallet's volume counts every trade it was part of. Shares are of the two-sided total (each trade counted once for the buyer and once for the seller), so they add up to 100%. Net is bought minus sold; taker flow is what takers bought minus what they sold. A big print is every fill one taker got in one block, shown as one order, from $100K.

Size is Hyperliquid's leaderboard account value, which includes spot collateral. For wallets not on the leaderboard we use our latest poll of their perp account; a $0 there counts as unknown, because accounts that keep collateral in spot read $0 when flat. Many market makers are on neither, so "size unknown" is its own group rather than being counted as small.

• Accounts of $10M or more
• $1M to $10M
• $100k to $1M
• Under $100k
• HLP: Hyperliquid’s own market-making vault and its sub-vaults
• Size unknown: on neither the leaderboard nor our position polls

Every wallet that traded at least $1M on the tracked markets in the last 7 days gets a type, recomputed hourly. A type says how the wallet trades, never who it is. It comes from four measurements of the wallet's own fills:

• Maker share: how much of its volume came from resting orders.
• Hours active: the share of all hours in which it traded at all.
• One-way share: for every market and hour, |bought − sold|, summed and divided by its volume. Near 0 when it buys and sells within the hour, 100% when every hour goes one way.
• Fill rate: fills per hour in which it traded.

The measurements form a continuum, not neat clusters, so the thresholds are judgement calls, set on the first days of data and published here. Every wallet page shows its own four numbers next to its type, with a confidence: high when each deciding number clears its threshold by a margin, with at least 500 fills and 48 hours of data; low with under 50 fills or under a day of data; medium otherwise.

One firm can look different on different wallets: a market maker's hedging wallet can be directional. And a wallet named after a market maker can still trade one way. When a label and a type disagree, both are shown.

Every label carries a tag for where it comes from, in this order of trust:

• ON-CHAINHyperliquid’s own records vouch for it: HLP and its sub-vaults (through the vault parent/child link), HIP-3 deployers and their fee wallets.
• REPORTEDPublic reporting ties the wallet to a named firm. Each one is reviewed by hand and listed below with its sources; the name never appears without them.
• COMMUNITYA community label from Hypurrscan’s shared alias list.
• RUNS VAULTThe wallet runs a Hyperliquid vault. That is a chain fact; the vault’s name is chosen by its creator.
• SELF-NAMEDA name the owner chose: leaderboard display names and vault names. Unverified, and they can impersonate.

Firm attributions are the only labels that name a company. One goes in only when a reputable outlet or on-chain analytics firm has published the wallet, or its truncated form, together with the firm's name. We link every source, say whether the firm has confirmed it (most never do), and never name private individuals, however well known. An attribution comes off when its source is retracted or the firm disputes it with evidence.

Lean is the long and short positions of the tracked wallets on a market at the latest poll, and how the net long share moved over the window. It covers the biggest wallets, not everyone, and positions can change between polls.

Steady traders are wallets that made at least $1M trading perps, and made it steadily. Hyperliquid's leaderboard ranks by total profit, which counts spot holdings rising in value; some of its biggest "earners" have no trading volume at all. Instead we read each wallet's own perp profit history from Hyperliquid (its portfolio record: cumulative perp profit since its first trade, about one point a week), which covers every perp market, trade.xyz included, and leaves spot out.

Who is checked: every wallet on the leaderboard with $1M+ total profit, $10M+ lifetime volume, some trading in the last 30 days and $100K+ in its account today; protocol vaults (HLP) are left out. The leaderboard folds sub-accounts into their main account, so when a main account's own perp volume is well short of its leaderboard volume, its sub-accounts' histories (up to 100) are added to its own. For one main account with 16 sub-accounts the sum came to $205.3B of volume against the leaderboard's $204.8B.

Steady means all four of: perp profit of at least $1M; at least 6 full 30-day periods of history, counted back from its latest point; money made in at least 60% of those periods; and a worst drawdown (the biggest fall of the curve from a previous peak) of at most 50% of all the profit it has made. The thresholds were set on a sample of the leaderboard's top earners; about one in four passed.

Records are rebuilt daily from 03:00 UTC, one request every few seconds so the live collection keeps the API budget; positions of every steady trader are then polled with the other tracked wallets. A fund whose perp position is one half of a hedge (the other half on another venue) has a perp curve that swings with the market; it fails the drawdown rule rather than showing as steady. A steady record says how a wallet has traded, not how it will.

• Who is behind most wallets. Market makers and funds rarely say, and one firm often trades from many wallets; each is counted separately.
• Hedges on other venues. A wallet that looks directional on Hyperliquid may be flat across all its venues.
• Orders that never traded: quotes, cancels and the order book itself are not part of these figures.
• Binance, Bybit and OKX volume by trader. Their totals appear in the venue panel; who trades there is not public.
• Anything before collection started, and markets we do not stream (the rest of Hyperliquid's list).