What this oracle ranking measures
Prediction markets price outcomes as probabilities. After a market resolves, you can ask whether a trader's entry prices were well calibrated — not merely whether they made money. Polydata's oracle leaderboard ranks wallets on resolved positions using Brier score (lower is better) and accuracy (share of positions that resolved in the trader's favor).
A random 50/50 baseline has Brier around 0.25. Strong predictors can sit well below that across hundreds of resolved books. Perfect 0.000 Brier with tiny volume is usually noise — always check positions and notional before treating a row as meaningful skill.
How to use the table
- Prefer wallets with both solid Brier and meaningful volume / position count.
- Open the oracle profile for a calibration curve (price buckets vs outcome rates).
- Cross-check PnL and Smart Score — accuracy without profitability (or vice versa) tells a different story.
- For live size, use the whale feed; for platform scale, see State of Polymarket.
Coverage and limits
Qualifying traders need at least 20 resolved positions in the selected window (all-time or 90 days). Shorter windows may be empty until backfill is available. Protocol / non-trader addresses are excluded. Figures come from Polydata's resolved-position index and can lag Polymarket UI. Wallet PnL methodology is documented on methodology; the broader product map is on Polymarket analytics.
Polydata never executes trades. Oracle ranks are research signals for investigation, not recommendations to copy anyone.