Loading your account

Ranking traders by profit does not find the best crypto traders to follow

In shortA profit ranking does not find the best crypto traders to follow. In the test behind our methodology note (2026-09-02), on wallets from our universe of 1,300+ Hyperliquid wallets, five of the top twenty by expected return had negative per-position expectancy on their own trades. What profit hides matters more: how deep a whale's worst case runs, and whether it holds long enough to follow.

If you searched for the best crypto traders to copy, you were most likely shown a ranking by profit: who made the most money, or the highest return, over the last week or month. We tested profit as a way to choose traders worth following, and it did not survive. Ranked by a profit-based model of expected return, five of the top twenty wallets in our test had negative per-position expectancy on their own trades.

The best crypto traders to follow are not found by sorting on profit. They are found by checking what profit hides: how large a position the trader's worst case allows, whether the trades a ranking scored are the trades an alert could deliver, and whether the trader holds long enough for anyone to follow. This page is the evidence for each, measured on Hyperliquid wallets, with the sample behind every number.

Why do lists of the best crypto traders rank by profit?

Profit is the easiest number to rank and the most satisfying to read. A ranking by dollars or by recent return is simple to compute, updates every day, and puts a dramatic figure at the top of the list.

A profit figure answers one question, who made money, and nothing else. It does not say how large the account was, how deep the trader went into drawdown to get there, which trades were counted, or whether those trades could have been followed in time. Each of those is a separate measurement, and each one changes who looks best.

Dollar profit also scales with account size: the same decisions on a bigger account produce a bigger number. That is why we measure every return on whole capital, spot and perpetual balances together, rather than on one venue's balance or as a dollar figure alone. Our guide to reading the Hyperliquid leaderboard covers the window and leverage traps; this page covers what happened when we tested profit as a selector.

What did we test, and on what?

We tested profit, and the rankings built on it, against what a follower can actually act on. The wallets came from a universe of 1,300+ Hyperliquid wallets that we evaluate on a rolling 90-day window, recomputed daily, as described in our methodology note of 2026-09-02.

TestSampleResult
Rank by a profit-based model of expected returnTop twenty by expected return, in the note's testFive had negative realized per-position expectancy on their own trades
Compare a tight and a wide worst caseTwo wallets from the same sampleA worst case 8.7× tighter, so identical risk buys 8.7× the position
Compare graded and emittable positionsFive live wallets, 90 days68.0% win rate on 247 graded positions, 56.0% on 468 emittable
Check whether a top-ranked wallet can be followedOne wallet ranked #2 of 20It closed 71.3% of its positions before a follower could act
Check which traits persistIndependent halves of the 90-day sampleWin rate +0.66, expectancy +0.34 (rank correlation)
Win rate against return187 whales with a full 30-day sample, 2026-10-02Spearman rank correlation 0.24

Every figure is built from the wallets' own trades, on their own whole accounts; none is a follower's result. The first five rows come from our methodology note, published 2026-09-02; the last was measured on our live board on 2026-10-02. Some samples are small, five wallets in one test and a single pair in another, and we say so beside each finding rather than rounding them into a headline.

Finding one: does profit tell you how well a whale can be followed?

No. How much of a whale's trade a follower can take is a sizing problem: a follower sizes a position against the whale's worst case, and neither the whale's profit nor its win rate enters that calculation.

The model we tested ranked wallets by expected return, built from each whale's returns and how tightly its positions could be sized against its worst drawdown. Five of the top twenty wallets by expected return had negative realized per-position expectancy: on their own trades, the average result per position was negative. A ranking led by returns put them near the top anyway.

The mechanism is sizing, not skill. The worst case we size against is the larger of a wallet's worst per-position drawdown and its worst realized loss on one trade, measured on whole capital, plus one point of headroom. Two wallets from the same sample show what it does:

Wallet AWallet B
Win rate52.4%78.1%
Whale monthly ROI35.2%109.2%
Worst case we size against (% of whole capital)1.089.37

Wallet B wins more often and made about three times the monthly return, so a profit ranking puts B first. Wallet A's worst case is 8.7× tighter, so identical risk buys 8.7× the position behind A on an entry of the same size. Win rate does not enter that calculation at all.

Two wallets are an illustration, not a study. The point they make is arithmetic: position size is derived from the worst case, so a ranking that ignores the worst case is ranking the wrong thing for anyone who sizes against it. None of this says what anyone following A or B would have earned; it describes position size per unit of risk, which is fixed before the trade and does not depend on how the trade turns out.

Finding two: are the trades a ranking grades the trades you can take?

Often they are not. In our own test, the ranking scored the positions it could grade, and only about half of what an alert could deliver was in that set.

Our selection graded positions that were reconciled, untruncated and held at least four hours, and all three conditions are knowable only when a position closes. An alert fires when a position opens. The set a follower could act on, the emittable positions, is therefore a different and larger set than the graded one.

Five live wallets, 90 daysPositionsWin rate
Graded (what selection scored)24768.0%
Emittable (what a follower could receive an alert for)46856.0%

Twelve points of win rate separate the two sets. Grading covered 52.8% of emittable positions, and an emitted signal landed in the graded set 48.1% of the time, a coin flip.

One wallet showed a far wider gap: 78.4% of 51 graded positions against 44.3% of the 149 a follower could have entered, over the same 90 days. The ranking we trusted did not survive contact with what a follower could actually take.

Finding three: can a top-ranked trader be followed at all?

Some cannot, and a ranking metric will not warn you. One wallet ranked #2 of 20 on our own ranking metric and closed 71.3% of its positions before a follower could act. Excellent, and un-followable.

A score built from closed positions sees a strong trader. The score cannot see that most of those trades were over before an alert could be read and acted on.

A ranking metric's blind spot for hold time is why a median hold of at least 2 hours is a gate in our selection rather than an input to a score. It is a copyability rule, not a quality rule: it says nothing about whether a whale is good, only whether its trades last long enough to be followed. For scale, across the live whales on our board on 2026-10-02, the median of each whale's median hold was 17.2 hours.

What predicts a whale worth following?

Of the traits we measured, win rate is the one that persists. Between independent halves of our 90-day sample, win rate had a rank correlation of +0.66, against +0.34 for expectancy, the average result per position and the closest thing to a profit-per-trade figure.

Persistence is what a selector needs. A trait that holds from one half of a sample to the other is likely to show up again; a trait that holds about half as well is partly a description of the period you happened to measure. On that test, profit per position was the weaker guide.

Persistence is not payoff, though: across 187 whales on our board on 2026-10-02, win rate's rank correlation with 30-day return was only 0.24. Win rate is the steadier trait and still a poor guide to return, which is why neither profit nor win rate is a ranking we would follow alone.

What do we look at instead of a profit ranking?

We look at risk before return: how deep a whale's worst case runs on its whole capital, whether its positions last long enough to follow, and whether its account is large enough for its ratios to mean anything.

Our own board is not exempt from the question this page asks, so we say it plainly: it lists whales by their 30-day realized dollars by default, because realized dollars are the plainest fact about what a whale did. That order is a display, not a verdict, and the colour printed beside every whale is the number to read first.

On 2026-10-02 each of the 199 live whales on our board carried a risk colour, and the colour comes from one number: the wallet's maximum drawdown per position on whole capital. Profit plays no part in it.

ColourMax drawdown per position, whole capitalLive whales on 2026-10-02
Whiteunder 5%60
Green5% to 10%32
Yellow10% to 15%25
Orange15% to 20%22
Red20% or more60

A colour is a statement about risk taken, not about returns coming. A white whale risks less of its account per position than a red one, and that is the entire claim; the bands are published on our how we measure page.

Two more checks sit beside the colour. The median-hold gate from finding three keeps out whales whose typical position ends before anyone can act on it. An account floor of $100,000 on whole capital, re-checked nightly, keeps out wallets whose returns are measured on almost nothing: we added it after two wallets on our board held almost nothing and a few thousand in profit, which made every ratio on their cards meaningless. A return on nothing is not a return.

How do you find the best crypto traders to follow?

Start from a ranking if you like, and then check what its sort hides. A ranking is a starting list, not a verdict.

  • What is it sorted by? Dollars favour large accounts, and short windows favour large bets. A rank is a place in an order somebody chose.
  • What is the worst case behind the return? Ask for drawdown per position on whole capital before you look at profit. Wallet B above had the better return and a worst case 8.7× wider than wallet A's, which decides size, not whether the trade pays.
  • Were the scored positions knowable when you would have acted? A figure built only from trades that passed a filter at close may describe trades no alert could deliver.
  • How long are positions held? If the median hold is shorter than the time it takes you to see and act on an alert, the ranking describes trades you cannot take.

What do these tests not show?

These findings have limits, and the limits belong beside the findings.

  • Small samples in places. The graded test used five wallets and the sizing example a single pair. They show mechanisms, not rates you can apply to every trader.
  • One market regime. Every threshold here was derived on a single 90-day window, and no walk-forward test across regimes exists yet.
  • Your result. These are whales' results on their own capital. A follower enters later, at a different size, and pays different fees, so we state no follower figure, and the exit is always yours: why copy traders lose money even when the trader wins shows why it decides the result.

Every figure above is a whale's own realized result on its own account, over closed trades only. Peak is the most a trade was up before it closed — nobody captures it, and it is never added to anything. A colour is a measured drawdown band, not a promise. How we measure.

Common questions

Who are the best crypto traders to copy?

No profit ranking can tell you. When we tested profit as a selector, five of the top twenty wallets by expected return had negative per-position expectancy on their own trades. Look instead for traders whose worst case per position is small relative to their whole capital, whose positions last long enough to follow, and whose figures come with their sample size and the date they were measured.

Is the most profitable trader the best one to follow?

Not necessarily. Profit does not enter the arithmetic that sizes a position. Two wallets from one sample differed 8.7× in the worst case a position is sized against, so identical risk buys 8.7× the position behind the tighter one, which was also the less profitable of the two. Our data does not show that the most profitable trader is the worst to follow either; it shows profit is the wrong selector.

Why don't profit leaderboards work for picking traders?

Profit hides three things a follower needs: how deep the worst case runs, which positions were counted, and whether the trades lasted long enough to follow. Across five wallets over 90 days, 247 graded positions won 68.0% against 56.0% for the 468 an alert could deliver, and one wallet ranked #2 of 20 on our own ranking metric closed 71.3% of its positions before a follower could act.

What should I rank traders by instead of profit?

Rank risk before return. We give whales a colour from one number, its maximum drawdown per position on whole capital, in bands from under 5% to 20% or more. Beside the colour we check that the median hold is at least 2 hours and that the account holds at least $100,000 on whole capital, so a return is not measured on almost nothing.

Can I check these numbers myself?

Yes. Every wallet on our board is a public Hyperliquid address with its own profile page, and Hyperliquid's data is on-chain, so nothing stops you recomputing any board figure yourself. Our methodology note lists the definitions and sample sizes behind each test, and every rate on our pages carries the number of closed positions behind it.

Everything above is measured from wallets you can check yourself on the Whales tab.

Whale data is public on-chain · not financial advice