About
What is Convexly?
Convexly is an independent intelligence and audit layer for prediction markets. For a Polymarket wallet it separates skill from luck: a four-state skill read, realized entry edge with a 95% confidence interval, the resolved-position count behind it, and a concentration check, on a reproducible, version-controlled methodology. No venue buys coverage and there is no fee-sharing with any venue; Convexly publishes its methodology and confidence intervals, and reports its negative results. It is built for prediction-market traders, prop desks, research teams, and data vendors that need a defensible, independent read on Polymarket trader cohorts.
What we publish
Open methodology. Our Edge Score composite has frozen coefficients, a published reference cohort, and validation scripts committed to git before the reported run. Code, limitations, and reproducibility details live in the V1 methodology paper.
Empirical studies. We publish the raw findings from our wallet studies, not just the conclusions. See the 10,000-wallet calibration audit.
Free tools. The Polymarket Wallet Analyzer and the Brier Score Calculator are public, no-private-key tools. Your first wallet analysis needs no signup, and the Brier Score Calculator needs none either; analyzing more wallets is free with an account. They are the same analysis layer we license to institutional buyers, just without the API and the custom reports.
How we publish
Methodology code is committed to a public git repository with pass-fail thresholds set before the validation script runs. Follow-up tests are filed externally with their hypothesis + analysis plan locked before any data is touched. Results publish whether they pass or fail; the negative-result registry at /research/negative-results is the receipt for that commitment.
Where a result ships with a public data bundle, it is reproducible from that bundle with the documented seed. Not every result does. The Polymarket cohort the V1 wallet-skill paper is fit on has never been published, so those results are recomputable by us and not yet by a reader; the V1-M Manifold half ships its cohort and its script. Corrected 2026-08-10, replacing a blanket claim that every quantitative claim was reader-reproducible. The audit chain walks SHA-256 hash-linked rows row-by-row in your browser at /research/verify. The full ex-ante filing index is at /research/preregistrations.
Contact
For methodology review, research collaboration, or institutional inquiries: research@convexly.app.
Common questions
Who is Convexly for?
Prediction-market traders, prop desks, research teams, and data vendors that need a defensible, independent read on Polymarket trader cohorts. The free Polymarket wallet check is the entry point; paid tiers add monitoring and cohort audits.
Is Convexly independent?
Yes. No venue buys coverage: no venue pays Convexly for a rating, reviews one before publication, or has editorial input, and there is no fee-sharing with any venue. Convexly does not take custody, broker trades, or recommend copying any wallet. Its methodology is version-controlled and its negative results are published. The full independence posture, including a dated statement of what Convexly has received from any venue to date, is stated at convexly.app/independence.
What does Convexly measure?
On a wallet's resolved Polymarket positions: calibration, position sizing, concentration, realized entry edge with a 95% confidence interval, and an Edge Score composite ranking. A record is reported as distinguishable from chance or not; Convexly does not predict a single wallet's future results and does not signal which wallet to copy.
Start with the work.