The numbers on a copy-trading leaderboard are real, and the exchanges publish the formulas. What is being sold is the sort order. Here is each field, what it is measured against, and the order to read them in — which is close to the reverse of the order they appear.
Open an exchange's copy-trading page and you are not looking at a database. You are looking at a database that somebody chose the sort order for. The default is nearly always a return figure over a short window, the cards that reach the top are the ones it flatters, and every other number is arranged around it.
The data underneath is real — the exchanges publish the formulas in public help pages. The pitch is not the data, it is the ordering. So here is each field: what it is measured against, what it cannot settle, and the order to read them in. The definitions below describe exchange documentation as it stood in August 2026; venues revise them, so check the help page of the platform you are on.
What is ROI actually measured against?
Not against your money. Against a capital base belonging to the leader, over a window somebody picked.
Binance documents its futures copy-trading ROI as PnL divided by MaxBaseBalance, times 100, where BaseBalance is the initial deposit plus net deposit and is refreshed every ten minutes; the stated purpose of the method is to stop ROI being eroded by frequent transfers (Binance, portfolio performance indicators in futures copy trading). OKX defines the equivalent field as "Lead trade PnL% = New lead trader total PnL / Investment during period", with the investment side adjusted for money moving in and out of the trading account, and it spells out a specific rule for that adjustment: when a day's transfers in are smaller than the cumulative net transfers out from earlier days, those transfers in are not added to the investment base at all (OKX, update to lead trader trading metrics).
Both definitions carry the same lesson: a percentage return is a ratio, and the exchange is telling you exactly what it put in the denominator. Two things follow.
The first is size. A denominator of a few hundred dollars turns ordinary dollar moves into extraordinary percentages. A leader on a small self-funded balance and one running a serious book can post the same headline while describing very different events, and nothing on the card adjusts for the difference.
The second is the window. Seven-day, thirty-day and ninety-day ROI are separate fields with separate values, and a page sorted on the shortest is sorted, in large part, on who was recently lucky. Nothing about a seven-day figure is dishonest — it is a measurement over seven days. It becomes a claim only when it sits at the top of a page as though it described a trader rather than a week.
In neither definition is the denominator your capital. What lands in your account is a different field, further down.
Why does total PnL grow with size rather than skill?
Because it is denominated in money.
Binance defines PnL as realized plus unrealized profit or loss, with the realized part taken net of trading fees, funding and insurance clearing fees. OKX derives its PnL from the change in account assets across the period with transfers backed out, and restricts the calculation to closed lead-trade orders.
Two things follow. First, a dollar total is a joint statement about skill and about capital, and you cannot separate the two by looking at it: sorting a leaderboard by PnL sorts it substantially by account size — a real fact about a trader, just not the one most people think they are reading.
Second, the Binance figure includes unrealized profit and loss, so an open position sits inside the headline while it is still absent from the win rate — two fields on the same card counting different moments of the same trade.
Whose profit is the follower profit?
This is the field most worth hunting for, and it is almost never the one the page is sorted on.
OKX publishes a "Copy trader PnL" on the lead trader profile, defined as "the sum of profits all copy traders made from copying the lead trader's trades" (OKX, lead trader profile) — a separate quantity from the leader's own, sitting on the same page. Binance's AUM definition implies the same split from the other direction: "Lead Trader's Portfolio Investment Amount + Total Copy Trading Investment Amount". Two pools, one of them the leader's own money.
The two diverge for mechanical reasons, none of them sinister. You enter when you press copy, not when the leader entered; your sizing, leverage, fees and any profit share are yours; and platforms cap how many copiers a leader can carry, with OKX showing current copy traders against a maximum capacity. What the leader lived through and what a follower lived through are two different series — the subject of the drawdown you inherit when you copy-trade.
One caution about the field: it is a sum, not a distribution. A positive total across all copiers can be carried by a handful of large accounts while the typical copier sits elsewhere, and the sum is taken over people who are still copying. Read it as the aggregate it is, and do not read a median out of it.
What does AUM tell you, and what does it not?
AUM is the field most often mistaken for a verdict.
OKX defines it as "the total value of assets currently being copied by traders copying a lead trader"; Binance defines it as the lead trader's own portfolio investment plus the total copy-trading investment. In both cases the bulk of it is other people's money, which makes AUM a measure of how convincing the page above it has already been — a popularity figure wearing an institutional word.
It carries one piece of genuine information, though not the one it is read for. A large AUM means the leader is now trading a substantially bigger book than the one that produced the history above it, and whether an approach survives that change of scale is a real question the leaderboard does not answer. The word "currently" matters too: AUM says nothing about when the money arrived, or how fast it is leaving.
Why is the length of the record the first thing to read?
Because it bounds what every other field on the card is allowed to mean.
The exchanges concede this in their own product design. Binance displays the Sharpe Ratio only for portfolios trading thirty days or more, refreshing it twice a day — the venue declines to risk-adjust a record shorter than a month. That is a useful standard to borrow: if the exchange will not risk-adjust a two-week history, you should not either.
There is a deeper problem no field can fix. A ranking contains only accounts that are currently on it. Profiles that stopped, blew up or were quietly abandoned do not rank low — they are absent. Nothing tells you whether the profile you are reading is a person's first, and a fresh profile shows a clean start rather than a first chapter. The visible history on any leaderboard is, structurally, a survivor's. That is not an accusation against any trader; it is a property of ranked lists.
How many closed trades it takes before a percentage means anything is a separate question with a real answer: see what sample size makes a hit rate real.
What is the win rate counting?
Something slightly different on every venue, which is the first thing to know.
Binance computes win rate as the number of profitable closed positions divided by the total number of positions, and its documentation is explicit that only completely closed positions count — partial closures do not. OKX uses profitable orders over total orders, and publishes a separate profit/loss ratio defined as winning orders over losing orders. OKX has also changed what feeds those fields: contract trades on a lead trader's selected lead-trade contracts now count even when risk-control limits stopped them becoming lead trades.
Three practical consequences. A trader who scales out of positions is measured on a different denominator from one who exits all at once, so the same discipline scores differently by exit style. An open loser stays out of the denominator until it is closed, so a win rate can hold steady while an unrealized loss grows beside it. And a win rate is not portable: the same trading, measured by two venues, produces two numbers — comparing them across platforms compares definitions, not traders.
The larger point is that the field decides very little on its own even when measured cleanly — the whole of why win rate alone means nothing.
What does the drawdown figure actually cover?
The worst thing that has already happened, inside a window somebody chose.
Binance defines maximum drawdown as "the maximum observed loss from the peak to the trough in the asset curve over a specific period", computed against the record's own historical ROI peak. It is a well-defined and genuinely useful quantity. It is also the most misread number on the card, for two reasons.
It is bounded by the record: a short history has had less opportunity to contain a bad event, so a small drawdown on a young profile is evidence of youth more than of risk control. And it describes falls already in the data — a decline in progress is only as deep as it has got so far.
Treat the published drawdown as a floor under your worst day, not a ceiling over it. The arithmetic of recovery belongs beside it: an account down 50% has to double to return to level, and one down 70% has to more than triple.
What one drawdown number cannot show is the shape of the path — one event or a slow bleed, recovery earned back or handed back, a single bad chapter or a pattern. That is what an equity curve is for, and it tends to take only a few shapes: five equity curve shapes and what each one is telling you.
In what order should you read the card?
The page presents these fields roughly in order of how persuasive they are. Read them in order of how hard they are to produce by accident, which is close to the reverse:
- Length of the record. It sets the error bars on everything else.
- Maximum drawdown, and the window it covers. The worst day already in the file.
- The shape of the equity curve, not its endpoint.
- Copier profit against the leader's own. The only field about people in your position.
- AUM and remaining capacity. How much bigger the book is now than the one that made the history.
- Win rate and profit/loss ratio, with the venue's definition in hand.
- Headline ROI. Last, and in the context of the six above.
Sorting the page by the field you intend to read last is the habit worth breaking. Sort however you like to gather candidates — just do not let the sort do the reading.
What can none of these fields prove?
One limitation applies to every field above, and no amount of careful reading gets around it.
A leaderboard measures one account. OKX's documentation is explicit that its lead-trading metrics isolate lead-trade orders on the trader's selected lead-trade contracts, excluding spot, bot, options and non-lead contract trading. That scope is a reasonable engineering decision with an unavoidable consequence: a position held anywhere else — another account, another venue, the spot book — is invisible to the numbers. A leader who is long here and hedged elsewhere renders identically to one who is simply long, and no field on the page separates them.
The remaining limits are the ordinary ones: the record does not carry forward, your fills are not the leader's fills, and a public metric describes what an account did rather than what a person will do next.
None of this argues against reading leaderboards. The fields are public, defined and mostly candid about what they measure — the exchanges write the formulas down. What is being sold is the arrangement, and that layer has begun to attract supervisory attention: the European Securities and Markets Authority published a supervisory briefing on copy-trading services on 30 March 2023 covering, among other things, information requirements on marketing communications and the qualifications of traders whose trades are being copied (ESMA).
Until that settles, the defence is the boring one, and it costs a few minutes per card. Read the definition before the number. Read the length of the record before the return. And treat the sort order as the one field on the page that was chosen for you rather than by you.