How we pick whales

Every whale is a real Hyperliquid account. The engine reads Hyperliquid's public data (its leaderboard, each account's fills and its account value) and nothing an account says about itself.

It starts from the leaderboard's accounts above a capital floor and keeps the ones that trade actively and do not look like market makers.

Each night the engine rebuilds every candidate's positions from its fills over the last 90 days and measures them. Five gates ask the questions below; the thresholds are on each whale's page, under Selection.

Closes, 90 daysenough closed positions in the last 90 days to rate the whale at all

≥ 50

Engine win ratethe share of its positions a follower could have taken that ended above zero after fees

≥ 60%

Worst concurrent drawdownhow deep its open positions went together, as a share of the whole account

< 50% of account

Median holdwhether its positions stay open long enough for a person to follow

≥ 2h

All-time ROIwhether the account is up over its lifetime: a sign test, nothing more

above zero

The board also needs a whole account of at least $100,000, re-checked every night.

Accounts that pass every gate rank first; the rest rank by the same score below them, and the board is filled in that order. A newcomer waits a day before it can take a place, and takes one only when there is room or when it beats the weakest whale on the board by a clear margin for more than one night running. A whale with no fills for 30 days is set aside; nobody is deleted, and a whale that rates again comes back.

By rule, a whale on the board stays until its account falls under the floor, its median hold falls under the Median hold gate above, or it loses money over 30 days with a low win rate or a large loss on its account, 3 nights running; then it is paused, and its page says why. A paused whale comes back after 3 clean nights running. One paused for losing money also needs a 30-day profit.

A whale's page shows its selection record: the day it was selected, the five gates with the whale's own values that day and the threshold under each, and every promotion, pause and return since.

Colour: the whale's worst single position in 90 days, as a share of its account, on the same thresholds for every whale. White is the smallest drop, red the largest. A whale with too little verified history to rate has no colour yet.

• whiteworst single-position drawdown in the last 90 days is under 5% of account
• greenworst single-position drawdown in the last 90 days is 5–10% of account
• yellowworst single-position drawdown in the last 90 days is 10–15% of account
• orangeworst single-position drawdown in the last 90 days is 15–20% of account
• redworst single-position drawdown in the last 90 days is 20% or more of account

Copy-grade is a stricter bar, checked every day. A whale is copy-grade when:

1. it passes the engine's whale filters that night;
2. its whole account is at least $100K;
3. it closed at least 20 positions in the last 30 days;
4. it made money over the last 30 days, and over the last 90;
5. it has enough measurable closes for a win rate (at least 5);
6. it does not trade like a bot: at most 30 open positions, and no more than 10 closes on a typical day;
7. no single close is half or more of its 30-day profit.

Copy-grade doesn't limit Copy or Mimic.

We promise no result. The whale's numbers are the whale's own: they are not your results, and past trades are not a forecast. A whale can change how it trades, pause, or leave the board.

We do not fill in a figure we could not measure: it is left out, never estimated.

Nothing in the app is a recommendation to trade.