Skip to main content
The prediction_markets schema is a cross-venue unified dataset that merges Polymarket and Kalshi onto one schema. Use it when you want to write a single query that returns results across both venues without per-venue UNIONs.
Maintained by: Dune · Refresh: ~1 hour · Venues: Polymarket (Polygon) + Kalshi

Available Tables

prediction_markets.markets

Unified market dimension across Polymarket and Kalshi: names, category, status, resolution and parlay legs

prediction_markets.trades

One row per taker fill across both venues with normalized P(Yes) pricing, side traded and taker action

prediction_markets.ohlcv_hourly

One Yes-side candle per market-hour across both venues, prices in [0, 1] probability space

prediction_markets.open_interest_hourly

Hourly open interest and traded volume per single market across both venues

Identifier conventions

Every row carries a venue column ('polymarket' or 'kalshi'). Identifiers are venue-scoped: market_id is a Polymarket condition_id (or combo condition ID for Polymarket Combos) when venue = 'polymarket' and a Kalshi ticker when venue = 'kalshi'. Always join on (venue, market_id). event_id and series_id follow the same rule: Kalshi’s event and series tickers, or, in the v2 tables, Polymarket’s own event and series ids. They are NULL where the venue has no such level (Polymarket Combos, markets Polymarket does not list, markets outside a series) rather than a synthesized ID.

Notes

  • markets is the dimension; trades, ohlcv_hourly and open_interest_hourly are facts. trades and ohlcv_hourly carry no market names or attributes beyond category, and open_interest_hourly carries none at all. Join markets on (venue, market_id) for labels, category, status and resolution.
  • Every market is presented from its Yes side. Yes is the venue’s positional side 0 (outcome_index = 0 on Polymarket), never a label heuristic. For team-vs-team markets check yes_outcome_name to see which side is Yes.
  • price in trades and OHLCV is always P(Yes) in [0, 1]. In trades, outcome_side says which side was traded, taker_side whether the taker bought or sold it, and taker_fill_price is the price of that contract, so taker_fill_price × num_contracts = amount_usd.
  • category is a unified enum (sports, crypto, politics, finance, technology, culture, weather, world, health, other), plus mixed for parlays whose legs span several categories. The raw upstream classification is preserved in markets.category_native.
  • Resolution uses one vocabulary everywhere: yes, no, void (refund) and scalar (fractional payout) in markets.resolved_outcome and ohlcv_hourly.market_outcome, which also reads unresolved before settlement.
  • Parlays are included from both venues: Kalshi parlays and Polymarket Combos (multi-leg parlays), flagged with is_parlay = TRUE. For like-for-like single-market comparisons, filter is_parlay = FALSE.