What the Polymarket Leaderboard Actually Measures

The Short Version

Polymarket's monthly PnL leaderboard ranks wallets by a figure that does not reconcile with the wallet's own all-time accounting. A month is a subset of all-time, so a monthly figure larger in magnitude than the all-time realized figure cannot both be true, absent an accounting boundary.

That boundary is the leading innocent explanation. Polymarket Perps has its own account and its own collateral, and the endpoint used here exposes no perps field. If the leaderboard counts perps PnL that the endpoint does not report, the gap is an accounting artifact rather than an error. That caveat limits what the rest of the study can claim.

26 of the 100 wallets earned more from rebates and refunds than from trading. For a copy trader, that changes what the wallet represents. A market maker with high trade counts and fee refunds is a different bet from a directional trader with an edge.

The Reconciliation Gap

Sample: the top 100 wallets on the monthly PnL leaderboard, collected September 13, 2026. Each monthly figure was compared with the same wallet's all-time realized market PnL from /v2/user-stats.

The 1x tier is inflated by marginal cases in the 2–6% range, which are most likely snapshot timing between the two endpoints. The 2x tier and above is the defensible signal. Aggregate for the sample: $70,889,247 claimed for one month against $54,878,533 of all-time realized PnL, or 1.29x.

The Rebate Economy

Non-trading income across the sample, all-time. These are income categories the endpoint reports separately from trading PnL.

Against $54,878,533 of realized trading PnL, non-trading income is 36.7% of the two combined. 26 of 100 wallets earned more from rebates and refunds than from trading. Fee refunds are heavily concentrated: the top five wallets hold 70.4% of all refunds in the sample, and those wallets run trade counts in the hundreds of thousands to millions. Those wallets behave like market makers.

What to Check Before Copying a Wallet

A leaderboard PnL that blends venues and income types is not a measure of prediction-market trading skill. If you are choosing a wallet to copy, evaluate the accounting rather than the headline.

• Sample size. Look for at least 100 trades over three months, not one hot week.
• Realized versus open PnL. A leaderboard figure can include open positions. Ask what is realized and what is still at risk.
• Fee and rebate dependence. If most income is refunds or rebates, you may be copying a market maker whose edge does not transfer to a follower.
• Drawdown and category consistency. A single category or a single large win can carry the number.
• Cross-check the wallet's own accounting. Compare the leaderboard figure with the wallet's all-time stats before committing capital.

BlockRotate's paper trading mode exists for this step: run a leader in simulation before you copy them with real capital. See the comparison of Polymarket bots for the tools and the criteria.

Method

• Source: Polymarket Data API v2, public endpoints.
• Sample: the top 100 wallets on the monthly PnL leaderboard, collected 2026-09-13T08:10:08Z.
• Reconciliation:each monthly PnL figure was compared with the same wallet's all-time realized market PnL from /v2/user-stats.
• Non-trading income: fees refunded, taker rebate, maker rebate, sponsored income, reward income, yield income, and referral income, summed across the sample.
• Reproducibility: the anonymized row-level dataset is downloadable below. Wallet identifiers are salted SHA-256 hashes; raw wallet IDs are available to researchers on request for verification.

Limitations

• Perps is the leading innocent explanation. Polymarket Perps is a documented product with its own account, its own collateral, and authenticated-only account data. /v2/user-statsexposes no perps field. If the leaderboard counts perps PnL that this endpoint does not report, the reconciliation gap is an accounting-boundary artifact, not an error. Do not read this study as a claim that any named wallet's figure is false.
• Single snapshot, single month. This is n=100 at one point in time, not a time series.
• Field interpretation.The endpoint's fields are used as named. Any inference beyond their documented meaning is stated as an inference.
• No named-wallet claims. The published dataset is anonymized. The aggregate findings are the claim.

Raw Data

The anonymized row-level dataset contains the monthly PnL, all-time realized PnL, ratio, trade count, volume, fees, and every non-trading income category for the 100 wallets in the sample.

Related

• Polymarket leaderboard guide
• Best Polymarket bots compared
• Polymarket copy trading with BlockRotate
• Is Polymarket legal? US and EU rules