How traders are picked
Who tandem reads each night, who is left out and why, what is scored, and what a pick's warnings mean.
tandem's picks in For You and Explore are not a leaderboard of the biggest profits. The top of a profit board is often one lucky week or one huge trade. Every night, tandem runs the same written rules over Hyperliquid wallets and keeps the ones a copier could have been glad to follow.
The question the rules ask is: would a retail trader be glad they copied this wallet with For You, starting today? Profit alone never answers it. A wallet worth picking:
- wins steadily, not on one lucky day or one lucky trade;
- cuts its losses at about the same place each time, and doesn't let losers run;
- has a real edge: it wins more often than its risk : reward needs;
- falls less than a copier's plan allows;
- trades at a size and pace a copy can follow, on markets For You copies;
- is active now.
A pick is a record, not advice
Everything here describes a trader's past results. A pick can lose money, and your copy enters later, at your own size and prices, with your own limits and costs. Losses can exceed your limits. See Risks.
Who is read
Every day Explore tracks about 5,000 Hyperliquid wallets, chosen by their profit over the last 30 days. The nightly research reads the ones among them that made at least $1,000 in perps over those 30 days, which has been about 4,000 wallets on recent nights. Each is judged from its own trades, as Hyperliquid records them: how it entered and exited, how it fell, how often it traded and what it held open.
Who is left out, and why
A wallet that trips any of these rules is not picked. Each rule is there because a copy could not follow that wallet, or because its record would mislead.
| Left out | Why |
|---|---|
| Vaults | A pooled fund, not a person's own trading. |
| Likely market makers | A wallet that mostly places resting orders on both sides of the book, many times a day, and ends up near flat. Its profit comes from quoting, which copying its positions does not reproduce. A wallet that a trader's page would call a likely market maker is never a pick. |
| Emptied accounts | A wallet with under $1,000 in its whole account is never picked: a copy needs money to split, and Hyperliquid's smallest order is $10. |
| Profit without real perp trading | A wallet with an empty perp account, profit but no perp trading, or trading spread through sub-accounts: its record does not show its trading. |
| Hedgers | A wallet whose positions are offset on spot. Its perp profit is not a bet a copier would hold. |
| Not trading | No trade in the last 30 days. |
| Lost money over 90 days | Its profit over 90 days is below zero. A wallet whose closed trades lost money over 30 or over 90 days is not picked either. |
| Liquidations | Three or more liquidated trades in 90 days, or one that took 10% or more of the account, or two that took 3% or more each. |
| Too fast to copy | A usual hold under a minute, or more than 300 trades a day. |
| Too few trades | Under 15 closed trades in 90 days (6 for a swing trader, who trades less often), once its record is long enough to tell. |
A wallet whose figures are not read yet waits for the next night. A missing figure is never counted as a pass.
What is scored
The wallets that pass are sorted by trading style: scalpers, day traders, swing traders and quants (see the Glossary). Inside each style they get a score from 0 to 100 from six parts. Each part averages the measures tandem knows, and a part counts for more or less:
| Part | Counts | What it looks at |
|---|---|---|
| Consistency | 3 | Weeks up, days up, and no single lucky day. |
| Edge | 3 | Whether it wins more often than its risk : reward needs, and what an average trade makes. |
| Risk | 3 | Its worst fall, how well it cuts losses, and the leverage it uses. |
| Copyability | 2 | Its Fit, how much money a copy needs, and how long it holds. |
| Return | 2 | Its return over 30 and over 90 days. A wild month can't buy a place: the return part is capped. |
| Recency | 1 | How recently it traded and on how many days. |
The wallets that score best in each style are then read closely, trade by trade: where each cuts losses and takes profit, which coins it trades and how deep their books are, whether most of the profit came from one coin or one week, and whether a run of losses is behind it. That reading also writes the short line on the card, in plain words and with the trader's own figures.
What the strongest picks show
A wallet that passes the rules above is still judged on each of these points:
- The same answer over 30 and over 90 days. It was profitable over both, and its 90-day return is known.
- A worst fall of 20% or less over 90 days, which is the Balanced plan's stop level, and typical leverage of 10× or less.
- It cuts its losses. A loss that was let run and cost over 5% of the whole account at its time keeps a wallet off the picks.
- No single day or trade carrying the profit. Its best day may not make most of the month's profit, and its best trade may not make most of its 90 days' profit. A trader that is already a pick gets a little more room on the best day, so one lucky day doesn't flip a trader off the list and back on.
- Enough losing trades to judge by. With fewer than 10 losing exits in 90 days its habits can't be read well enough yet.
- A copy can follow it. At most 30 trades a day, a 30-day return of at least 3%, and a trade in the last 7 days.
- No liquidation of 3% of the account or more in 90 days, and a full 90 days of history.
- Not a fresh pump. Most of the 90 days' profit made in the last 30 days, with few weeks up, holds it back, and so does profit made elsewhere than in its own perp trading.
A wallet that shows all of these is a tandem pick. One that passes the rules but misses one point, such as a short history or a single large trade, is held back a step: For You marks it tandem watching, and its hover says "our research screened this trader and looked closely at its trades. We keep watching before we pick it." Each style keeps its lists short, at most 8 picks and 15 watched, because a few honest picks beat many forced ones. Where scores are close, the list is spread across account sizes and main coins, so a style isn't eight BTC whales.
How often it changes
The list is made again every night from that day's figures. A wallet stays a pick only while it keeps passing. Between nightly runs, a daily check takes a pick off the list if it has fallen apart, for example when its account has emptied. For You also holds back the Copy button on a pick whose account is under $1,000 now, or that has a position within 10% of its liquidation, and says why.
The page says when the research last ran, in a line that starts "Picked".
What a pick's warnings mean
A pick can still carry an amber warning. A warning says what to watch, it does not remove the pick. Each is about the trader's own record over the period you chose, never a promise of what your copy would make.
| Warning | What it means |
|---|---|
| Fell X% at its worst | Its worst fall is large against your plan: more than 80% of the drop your plan's stop level allows. |
| One day made most of its month | Its best day made over half of its profit over the period. |
| Small month: +X% or Small return | Its return over the period is under 3%. |
| Only N trades in 30 days | Fewer than 10 closed trades, which is little to judge by. |
| Quiet for N days | No trade for 7 to 14 days. |
| Trading for N days or Down $X since it started | Under 90 days of record, or no profit over its whole history. |
| No stops | None of its open positions has a stop order. |
| N% from liquidation | An open position is within 10% of being liquidated; the warning gives the distance. Copy is held back while it lasts. |
| Only $X in its account | Its account holds under $1,000 now, so a copy would copy nothing. Copy is held back. |
Two more things the research watches never keep a trader off the list on their own: a worst day in the last 30 that lost more than 5% of the trader's account, and a usual trade too small to pay after a copy's costs.