Big PnL is not the same as reliable behavior.
We score 8,656 Polymarket wallets and 15,106 Manifold users against frozen-coefficient, reproducible methodology. Kalshi trade-level data is in the V1-M paper. Vote on what we ship next.
What’s new
The latest features and improvements shipped to Convexly. We update this list every time something meaningful goes live.
- New
V1-M paper: cross-venue invariance
18,105,158 Manifold bets, 15,106 users, and one paired-window sweepcash comparison. The original bundle reported -8.9pp lower concentration; a 2026-05-04 recovered-cohort politics-excluded rerun reports -9.2pp (95% CI [-12.8, -3.6], p = 0.0021) on n=515 of 1,208 paired users with defined concentration delta. Calibration secondary claim superseded.
Try it - New
Wallet Ratings leaderboard (daily refresh)
Top-50 Polymarket wallets (≥20 resolved positions) ranked by Edge Score V3b. Refreshed every morning at 06:00 UTC by a scheduled pipeline. Current labels stay source-attributed to the public Polymarket cohort until partner data rights expand coverage.
Try it - New
Moat builder: multi-venue data-rights lane
The next coverage lane prepares Kalshi, Arkham, Robinhood, Gemini, and other venue integrations without claiming partner-only wallet identity data before access is approved. Labels remain provisional until the evidence source is explicit.
Try it - New
Public data bundle (V1-M)
15,106 hashed users, sweepcash within-user deltas, all refit coefficients with bootstrap 95% CIs and permutation null p-values, and a standalone reproduce.py that replays the analysis from the Manifold /v0 API. Released under the same terms as the Manifold public data policy.
Try it - New
Polymarket Wallet Analyzer
Paste any Polymarket wallet and get calibration, position sizing diagnostics, and concentration share in about 30 seconds. Category breakdown included. Free, no signup.
Try it - New
Position Sizing Diagnostics
Our 8,656-wallet Polymarket study plus the 15,106-user Manifold V1-M cohort show calibration alone barely predicts profit. Sizing and concentration dominate. See your category-level sizing pattern against the implied-odds reference.
Try it - New
Category Calibration
Brier scores by domain: politics, sports, crypto, economics. Know exactly where you are well-calibrated and where you are not. Where your calibration is strong, size up. Where it is weak, sit out.
Try it - New
Kalshi Market Context
Kalshi signed orderbook freshness is live for market-quality context. Account-level scoring remains user-authorized or partner-gated.
Try it Polymarket auto-resolution
Polymarket trades automatically resolve as their markets close on-chain. A weekly sweep keeps your calibration curve up to date. Researcher tier also unlocks manual syncs on demand.
Try itPolymarket wallet auto-import
Paste your Polygon wallet address and every filled trade becomes a Convexly decision. No credentials, no signatures, up to 2,000 trades per wallet.
Try itCalibration Challenge quiz
A 10-question forecasting test that returns your Brier score and calibration curve in 2 minutes. No signup required to try it.
Try it
Want to see what’s coming next? Vote on the roadmap below.
Planned Features
Sorted by votes
Chrome Extension
See wallet scores, market caveats, and sizing diagnostics where you research markets. Planned after the core analyzer, leaderboard, and evidence surfaces are tighter.
Unified Cross-Platform View
Planned venue-specific calibration and sizing profiles across Polymarket and Kalshi once each source clears its own evidence gate.
Prediction Review
When a market resolves, see where your probability beat the market and where it didn't. Were you right but undersized? Wrong but hedged? Pattern recognition on your misses.
Concentration Alerts
Get warned when your portfolio is overexposed to one event. V1-M's corrected read is narrower but still useful: concentration was lower in the sweepcash window, and the politics-excluded recovered-cohort rerun reports -9.2pp (95% CI [-12.8, -3.6], p = 0.0021) on n=515 of 1,208 paired users with defined concentration delta.
Suggest a feature
Don't see what you need? Tell us. We read every submission.