P&L is not skill: what the Polymarket leaderboard ranks

Polymarket's leaderboard ranks realized profit and trading volume, and nothing else. Its public API accepts exactly two ordering criteria, PNL and VOL, and the response carries no accuracy field, no hit rate and no count of resolved markets. That makes it an accurate answer to "who made the most money here" and a poor answer to "who knows what is going to happen".

What does the Polymarket leaderboard actually rank?

The documentation settles it. The public endpoint is GET https://data-api.polymarket.com/v1/leaderboard, and as of September 2026 its parameters and response schema are: category across eleven market categories, timePeriod of DAY, WEEK, MONTH or ALL, and orderBy with two allowed values, PNL and VOL. Each row returns rank, proxyWallet, userName, vol, pnl, profileImage, xUsername and verifiedBadge.

Read the response fields again. There is no field for how many markets the trader resolved, how often they were right, or how their entry price compared to the final outcome. Those numbers are not hidden behind a paid tier. They are not part of the ranking at all, because the ranking is not about accuracy.

This is not a criticism of Polymarket. A trading venue ranking traders by profit is doing exactly the sensible thing. The mistake happens downstream, when a profit ranking gets read as a skill ranking by people looking for someone to follow.

What does a profit leaderboard measure?

Profit is the product of three things: how right you were, how much you staked, and how good the odds were when you got in. Only the first is skill. The other two are bankroll and timing, and they dominate the arithmetic.

Consider two traders in the same market. One is right on 30 of 40 positions, staking $50 each, and finishes ahead by a few hundred dollars. The other is right on 3 of 10 positions, but one of those three was a $2 million position at 20 cents, and finishes millions ahead. The leaderboard has an unambiguous winner and it is the second trader. Every additional zero on the stake moves someone up the ranking without changing a single opinion.

Formal forecasting research fixed this problem seventy years ago by scoring the probability rather than the payout. The IARPA tournaments run from 2011 to 2014 scored every participant with the Brier score, which is simply the squared distance between the probability you gave and what actually happened, so that a confident wrong answer costs more than a hedged one and stake size never enters the calculation. Selecting on that measure, the winning program identified 60 superforecasters purely on accuracy and watched them stay ahead of comparison groups for two further years. No version of that result is reachable from a profit column.

Four ways to top a profit leaderboard without being right

Deploy more money. Two people with identical opinions and a 10x difference in bankroll produce a 10x difference in profit. The leaderboard sees only the second number.

Take one large directional position. A single big call that lands outranks a season of small correct ones. The 2024 US presidential election is the clearest case on the record: The Block reported on 7 November 2024 that one French trader had spent more than $45 million on Trump positions and finished with an estimated $78.7 million in net profit. Whether the underlying analysis was good is a separate question from the ranking, which would look identical if the same money had been placed on a hunch.

Farm favorites. Repeatedly buying outcomes at 95 cents produces a long string of wins and a slowly growing profit line. It is close to riskless and it is close to informationless, because agreeing with a 95% price adds nothing that the price did not already say.

Use several wallets. Rankings are per wallet, not per person. Chainalysis linked nine separate Polymarket accounts to that single French trader, identified through funding patterns, transaction timing and cash outs to the same exchange deposit addresses. One person, nine rows.

Why is volume an even weaker signal?

Volume counts dollars traded, win or lose. A trader who loses steadily while trading constantly climbs a volume ranking. That alone should disqualify it as a proxy for skill, and there is a further problem.

In October 2024, two blockchain analytics firms examined Polymarket's election activity for a Fortune investigation. Chaos Labs concluded that around one third of trading volume on the presidential market showed signs of wash trading, the practice of buying and selling with yourself to inflate apparent activity. Inca Digital separately found that on chain transaction volume for the 2024 election market came to about $1.75 billion against a reported figure of about $2.7 billion. Whatever the right final number, a metric that can be manufactured by trading with yourself is not measuring insight.

Why do people follow these rankings anyway?

Because a leaderboard looks like a verdict, and because almost every site publishing one has a reason to say it means something. Copy trading tools, whale trackers and alert services all need the premise that the top of a profit ranking is a signal worth paying for.

There is a plainer reading. A profit ranking with no denominator is the same object as a track record with the losses removed: a numerator presented as a rate. CXO Advisory Group graded 6,582 public US market forecasts from 68 named experts between 2005 and 2012 and found 46.9% terminal accuracy, which is what a group of prominent forecasters looks like when nobody chooses which calls count. A leaderboard sorted by dollars never runs that test on anyone.

There is also a timing problem that no ranking can fix. By the time a large position is visible on chain, the price has already absorbed it. A follower buys the same outcome at a worse price, with none of the reasoning and none of the exit plan.

What does a skill leaderboard look like instead?

It measures the same three things a forecasting tournament measures, adapted to a live market: how often you beat the price, by how much, and how early.

Beats market. The share of your resolved calls that scored better than the market price at the moment you made the call. Being right about a 92% favorite barely moves it. Being right where the price said 30% moves it a lot.

Edge. How much better than the price your calls were on average, across everything you called. It is an average over all resolved calls, so a single spectacular call cannot carry it.

Early. Whether the market price moved toward your call after you made it and before it went public. This is the hardest column to manufacture, because it requires the rest of the market to react after you rather than before.

Events. The denominator. How many of your calls have resolved. Without it the other three columns are unreadable, which is why people under 30 resolved events are not ranked at all.

Column or field	Public in the API	Changes if the same positions are taken with 10x the stake	Changes if the person is right more often	Minimum resolved events required
Polymarket pnl	Yes	Yes	Yes	None
Polymarket vol	Yes	Yes	No	None
Polymarket verifiedBadge	Yes	No	No	None
insiderz Beats market	Yes	No	Yes	30
insiderz Edge	Yes	No	Yes	30
insiderz Early	Yes	No	Yes	30

Checked 4 September 2026 against Polymarket's published API reference and the insiderz leaderboard.

How do you read the insiderz columns?

Start with Events, because it decides whether the rest is noise. Then read Beats market as the honest hit rate, Edge as the size of the advantage, and Early as evidence about timing. A person with a high Beats market and a low Edge is right often on easy calls. A person with a moderate Beats market and a high Edge is right on the ones that were priced wrong. Both are useful, and they are different people.

Here are the top insiders right now.

Two limits worth stating plainly. insiderz launched in September 2026, so nobody on it has a long history yet, ours included, and a leaderboard from a young platform should be read as a start rather than a verdict. And no ranking, on any site, tells you what someone will do next. What a scored record can do is rule out the explanations that a profit ranking leaves wide open: money, luck on one position, and a hidden denominator.

Nothing on insiderz is staked, so there is nothing to copy in the trading sense. What you can do instead is read a person's record, follow them, and ask for live access to see their calls the moment they are made rather than after the delay. For the practical version of that, see how to find people who actually beat the market, and for what a record has to contain before any ranking on it means anything, see how to build a prediction track record people can verify. The events list shows the same markets Polymarket lists, priced the same way, with no money involved.