Prediction Markets

Polymarket, Kalshi, and cross-platform prediction market data.

17 tables across Polymarket (on-chain, Polygon) and Kalshi (centralized).

Polymarket Tables

Table	What it is
`agent.polymarket_trades`	Individual trade events (934M rows)
`agent.polymarket_market_details`	Market metadata (question, status, tags, links)
`agent.polymarket_events`	Event-level metadata grouping related markets
`agent.polymarket_prices_daily`	Daily outcome token prices
`agent.polymarket_prices_hourly`	Hourly CLOB mid-prices
`agent.polymarket_prices_latest`	Latest price snapshot per condition
`agent.polymarket_volume_daily`	Pre-aggregated daily volume per market
`agent.polymarket_volume_hourly`	Pre-aggregated hourly volume
`agent.polymarket_rolling_metrics`	Pre-computed rolling window metrics
`agent.polymarket_open_interest_daily`	Cumulative OI (sparse — see below)
`agent.polymarket_user_activity`	Per-user trade activity
`agent.polymarket_positions`	Current wallet positions with PnL

Kalshi Tables

Table	What it is
`agent.kalshi_market_details`	Market metadata (55.6M rows of snapshots)
`agent.kalshi_trades`	Individual trade fills (ReplacingMergeTree — use FINAL)

Cross-Platform Tables

Table	What it is
`agent.prediction_markets_daily`	Unified daily volume/OI across both platforms
`agent.prediction_markets_matched_markets`	LLM-matched equivalent markets across platforms
`agent.prediction_markets_matched_daily`	Daily comparison for matched pairs

Key Concepts

Join key: condition_id links all Polymarket tables together. market_ticker links Kalshi tables.

Volume: always one-sided (Paradigm methodology). Each OrderFilled event counted once with account=maker.

Prices: represent implied probability (0–1 range, where 0.65 = 65% chance).

Market deduplication: polymarket_market_details has two rows per condition (one per outcome). Filter outcome_index = 0 for one row per market.

Polymarket OI is Sparse

polymarket_open_interest_daily only has rows on days with OI-changing events (splits, merges, redemptions). To get the latest OI:

SELECT condition_id,
       argMax(open_interest_usd, block_date) AS latest_oi
FROM agent.polymarket_open_interest_daily
GROUP BY condition_id

Do not filter HAVING latest_oi > 0 — neg-risk markets have legitimately negative OI on individual conditions.

Good to Know

Polymarket data lag: ~1 hour. Kalshi lag: ~6 hours.
Kalshi hierarchy: series → events → markets. MVE (parlay) markets skip the event layer.
prediction_markets_daily only includes dates where both platforms have data.
Matched market pairs have confidence scores (75–100). Filter 90+ for highest quality.
Kalshi trades are anonymous (no wallet attribution). Polymarket uses proxy wallet addresses.