Best Polymarket Wallets to Study

The best Polymarket wallets to study are not the top of the realized PnL leaderboard. They are the top of the Edge Score leaderboard, where cross-sectional behavioral skill is easier to inspect. Across the full 8,656-wallet Polymarket cohort, calibration and realized PnL correlate at just Spearman +0.148. Edge Score V3b, a frozen-coefficient composite of posture, conviction, and discipline, correlates with signed log PnL at out-of-fold Spearman +0.514 (Fama- French permutation p less than 0.0001 on 10,000 shuffles). The full methodology is open at convexly.app/research/edge-score-methodology-v1.

Why the PnL leaderboard is a bad skill filter

Polymarket ranks wallets by realized PnL. That is the natural thing to rank by and it is the wrong filter for skill ranking, for three reasons that compound.

First, a single concentrated bet can produce a full career of PnL for a disciplined whale. If the trade has already resolved, the reader sees the wallet after the edge is gone. If the trade has not resolved, the reader is studying one event, not a skillset.

Second, realized PnL is conditional on survivorship. Wallets that took concentrated risks and lost are not at the top of the leaderboard. The leaderboard selects for the winning tail of a variance process. A reader who studies only today's winning tail has no reason to assume that tail repeats tomorrow.

Third, Hill alpha on per-wallet PnL in our V1 cohort is 1.28 (95% bootstrap CI 1.20, 1.36). That is meaningfully below equities at ~3.5 and borderline on whether even the population mean of per-wallet PnL is well-defined. Rank-based filters are less sensitive to this tail shape. Point estimates of “top wallets by PnL” are not.

What Edge Score filters for

Edge Score V3b combines three z-scored pillars fit against signed log PnL on the frozen 2026-04-15 cohort:

• Posture: the wallet's skill-Brier (Brier score net of baseline-frequency naive prediction). Positive posture rewards wallets that earn profit while being calibrated above their own baseline.
• Conviction: log of single-event concentration share. High conviction means most of the PnL sits in one event, which the fitted coefficient (+2.72) rewards on Polymarket V1 but should be treated with caution for turning the score into a trade-following rule.
• Discipline: negative log of resolved position count. The fitted coefficient is minus 1.15 on Polymarket V1 (more resolved positions predicts lower signed log PnL on the survivorship-selected cohort) but positive on Manifold play-money (per V1-M). Interpret per venue.

How to screen for a wallet worth studying

Three filters, applied in order, to any candidate 0x address:

1. Edge Score percentile above 80. Top 20% against the 8,656-wallet cohort.
2. Balanced pillar breakdown. A wallet driven entirely by conviction is a one- event track record. Prefer a profile where posture and discipline contribute meaningfully alongside conviction.
3. Concentration share below 0.40. The single-largest-event PnL share. Above 0.80 is effectively a one-bet wallet; under 0.40 implies the record is less dominated by one resolved event.

A wallet that clears all three filters is a reasonable starting candidate for further research. None of this is financial advice, and past skill does not imply forward returns, especially on a venue where Hill alpha sits near 1.

Does Convexly publish a recommended-wallet list?

No. We publish the methodology, the coefficients, the cohort snapshot, and the ranking itself (top 100 refreshed daily). We do not publish a specific recommendation list. The reason is narrow: recommending a wallet would require a forward-performance claim, and the V1 paper's limitations section explicitly flags concentration and survivorship as forward-claim risks. The correct primitives are the PnL-vs-Edge comparison (comparison view), the Polymarket wallet analyzer (per-wallet lookup), and the V1 methodology paper (the ex-ante rules). Build your own screen; the data is there.