prediction_markets.trades table holds Polymarket and Kalshi trades in one table. It has one row per taker fill, unique on (venue, block_month, trade_id). The taker is the side whose order filled against a resting order. Polymarket Combos (multi-leg parlays) are included; they have is_parlay = TRUE in markets.
Each trade appears once, so SUM(amount_usd) gives venue volume without extra filters. Polymarket’s maker rows are only in the venue table, polymarket_polygon.market_trades, where is_taker_side = FALSE.
Three columns describe each trade:
priceis always the probability of Yes in[0, 1], comparable across venues and sides.outcome_sideis the side of the market the traded contract belongs to (yesorno).taker_sideis whether the taker bought or sold that contract (BUYorSELL).
taker_fill_price is the price of the contract actually traded (price for a Yes contract, 1 - price for a No contract), so taker_fill_price × num_contracts = amount_usd on every row.
On Polymarket rows, event_id is Polymarket’s own event id, the same value as prediction_markets.markets.event_id.
Kalshi reports every trade as a purchase of the side the taker ends up holding. So
taker_side is always BUY on Kalshi, and closing a Yes position shows as a BUY with outcome_side = 'no'. Kalshi also reports one row per taker-maker pair. Compare sums across venues, not row counts.Table Schema
Table sample
Query performance
The table is partitioned byblock_month. Always filter on block_month or executed_at.
Example queries
Previous version
Previous version (v1)
Previous version (v1)
Columns of the previous version, for queries that still use it. The migration guide maps each old column to its new name.