Methodology

All data is pulled from public Polymarket and Kalshi APIs. No private feeds, insider access, or paid partnerships.

• data-api.polymarket.com/v1/leaderboard — top wallets by all-time PnL
• data-api.polymarket.com/positions?user=<wallet> — currently-held positions per whale
• data-api.polymarket.com/trades — recent fills used for the "today\u2019s picks" column
• api.elections.kalshi.com/trade-api/v2/markets/trades — Kalshi trade tape

Every endpoint includes last_updated and a 1\u20136 minute cache TTL. We never persist a private user identifier.

We use Polymarket\u2019s own leaderboard, ranked by all-time realized PnL. We do not apply our own scoring on top — the ranking is what any visitor sees on polymarket.com/leaderboard.

For the "Today\u2019s Leaders" tab we re-rank by period-volume so dormant historical winners don't crowd out actively-trading whales.

For every whale in the cohort we fetch their currently-held positions and group them by (market_id, outcome). We rank groups by count of distinct whales agreeing; ties broken by total $ at risk.

Single-whale groups are filtered out — we want shared conviction, not one mega-bet. Code: /app/lib/services/consensus.js. Server-cached 5 min.

Four metrics on the track record page:

• Aggregate ROI — sum of all-time realized PnL ÷ sum of all-time volume. Capital efficiency, not hit rate.
• Median whale ROI — median of per-whale (PnL ÷ Volume). Robust to outliers.
• Open-position win rate — fraction of currently-held cohort positions with positive unrealized PnL. Updated every 6h.
• Aggregate open PnL % — weighted-average percent PnL across all open positions, weighted by initial $ at risk.

A consensus track record with a true hit-rate and calibration curve goes live once we\u2019ve accumulated ≥30 days of snapshots. Live since: June 2025.

We disclose these openly because no methodology is bias-free:

• Survivorship bias — the "top 50" are by definition historical winners. Not predictive at the individual level.
• Multi-wallet operators — sophisticated traders sometimes split capital across wallets to obscure size. We dedupe by address only.
• Recency / lookback bias — "all-time" PnL is measured since each wallet started trading.
• Open-position win rate is unrealized — +12% today can resolve at zero tomorrow.
• No execution adjustment — weighted avg entry price is what whales paid; orderbook depth is shallow, you may not get the same fill.
• Sample size is small — 50 whales × ~5,500 open positions is one cross-section; long-tail markets often have 1\u20132 whales.

• We do not custody, hold, or exchange crypto.
• We do not place trades, execute orders, or accept bets.
• We do not issue financial advice, predictions, or recommendations.
• We do not guarantee profitability — see disclaimer.

Methodology is versioned. Current: v1 (June 2025). Material changes (new metrics, redefinitions, cohort size) logged here with effective date.

• v1.0 — June 2025: initial publication. Top 50 cohort, pure whale-count consensus ranking, 4 published metrics.