Prediction Markets

Polymarket, Kalshi, and cross-platform prediction market data — trades, market details, prices, volume, and open interest.

17 tables in this category.

Tables

View Name	Database	Source	ORDER BY
agent.polymarket_trades	agent	polymarket_polygon.market_trades	block_time, tx_hash, evt_index
agent.polymarket_market_details	agent	polymarket_polygon.market_details	condition_id, token_id
agent.polymarket_events	agent	polymarket_polygon.events	—
agent.polymarket_prices_daily	agent	polymarket_polygon.market_prices_daily	—
agent.polymarket_prices_hourly	agent	polymarket_polygon.market_prices_hourly	—
agent.polymarket_prices_latest	agent	polymarket_polygon.market_prices_latest	token_id
agent.polymarket_volume_daily	agent	polymarket_polygon.market_volume_daily	condition_id, block_date
agent.polymarket_volume_hourly	agent	polymarket_polygon.market_volume_hourly	condition_id, block_hour
agent.polymarket_rolling_metrics	agent	polymarket_polygon.market_rolling_metrics	—
agent.polymarket_open_interest_daily	agent	polymarket_polygon.market_open_interest_daily	condition_id, block_date
agent.polymarket_user_activity	agent	polymarket_polygon.user_activity_v2	account, token_id, block_time
agent.kalshi_market_details	agent	kalshi.market_details	market_ticker
agent.kalshi_trades	agent	kalshi.trades	ticker, created_time, trade_id
agent.prediction_markets_daily	agent	prediction_markets.daily	—
agent.prediction_markets_matched_markets	agent	prediction_markets.matched_markets	—
agent.prediction_markets_matched_daily	agent	prediction_markets.matched_daily	—
agent.polymarket_positions	agent	None	—

Related Tables

• agent.kalshi_daily_categorized
• agent.kalshi_market_report
• agent.prediction_category_daily

Sample Queries

1. Daily volume by platform

SELECT date, source,
       sum(notional_volume_usd) AS volume,
       sum(open_interest_usd) AS oi
FROM agent.prediction_markets_daily
WHERE date >= today() - 30
GROUP BY date, source
ORDER BY date DESC, source

Table Schemas

agent.polymarket_trades

Individual Polymarket trade events (OrderFilled) with USD amounts, maker/taker addresses, condition IDs, and outcome context. One row per trade event. The primary source for Polymarket volume and trading activity analysis.

Engine: MergeTree | Partition: toYYYYMM(block_date) | ORDER BY: block_time, tx_hash, evt_index

Column	Type	Description
block_date	Date	Partition key. Always filter on this.
block_time	DateTime64(3)	Trade timestamp.
block_number	UInt64	Block number
tx_hash	String	Transaction hash
evt_index	UInt32	Event log index in transaction
exchange_address	String
protocol	String
project	String	Protocol name
order_hash	String
maker_address	String	Maker (liquidity provider) address.
taker_address	String	Taker (liquidity consumer) address.
maker_asset_id	String
taker_asset_id	String
maker_amount_filled	UInt256
taker_amount_filled	UInt256
fee_raw	UInt256
outcome_token_id	String
shares	Float64
amount_usd	Float64	Trade notional value in USD (one-sided).
price	Float64	Token price in USD
fee_usd	Float64
condition_id	String	Unique market condition identifier. Links to market_details.
question	String
outcome_index	UInt8
outcome_label	String
category	String	Market category
market_slug	String
market_end_date	DateTime
neg_risk	UInt8
question_id	String
event_slug	String
event_title	String	Event title

• account = maker captures both sides of a trade — do NOT fan out to both maker and taker
• This follows Paradigm methodology: each OrderFilled with account=maker is one-sided volume
• 934M rows — always filter on block_date or block_time ranges
• condition_id links to market_details for market question and metadata
• JOIN with polymarket_market_details ON condition_id for market enrichment

• On-chain trades only — off-chain order book and cancellations not included
• Pipeline lag: Polygon BQ export runs every 1h, typical lag ~1 hour
• No aggregator/frontend attribution — all trades appear as CTF Exchange events

agent.polymarket_market_details

Polymarket market metadata dimension table. Contains market questions, descriptions, event groupings, resolution status, end dates, tags, and Polymarket links. One row per (condition_id, outcome_index) — use outcome_index=0 for one row per condition.

Engine: MergeTree | ORDER BY: condition_id, token_id

Column	Type	Description
condition_id	String	Primary market identifier. Links to all other Polymarket tables.
question_id	String	Groups related conditions into a single question (e.g., ‘Will X win?’ has Yes/No conditions).
question	String	Market question text (e.g., ‘Will Bitcoin reach $100K by December 2026?’).
market_description	String	Longer market description with resolution criteria.
token_id	String
outcome_index	UInt8	0 or 1. Use 0 for deduplication (one row per condition).
outcome_label	String
token_outcome_name	String
category	String	Market category
market_slug	String
event_slug	String	URL-safe event identifier for Polymarket links.
event_title	String	Parent event title grouping multiple markets.
market_end_date	DateTime	Scheduled market end date.
neg_risk	UInt8	Boolean. True = this is a neg-risk (multi-outcome) market.
active	UInt8	Boolean. True = market is currently open for trading.
closed	UInt8	Boolean. True = market has been closed (trading halted).
archived	UInt8
volume_total	Float64
liquidity	Float64
resolved_at	DateTime64(3)	Resolution timestamp. Null if unresolved.
created_at	DateTime
updated_at	DateTime
market_id	UInt64
start_time	DateTime
close_time	DateTime
game_start_time	DateTime
tags	String	Array of category tags.
volume_1wk	Float64
volume_1mo	Float64
volume_1yr	Float64
resolution_source	String
image	String
icon	String
neg_risk_market_id	String
event_description	String
event_start_time	DateTime
event_image	String
event_resolution_source	String
status	String	Market status (active, closed, resolved)
winning_outcome_index	Int8
polymarket_link	String	Direct URL to this market on polymarket.com.

• Filter outcome_index = 0 for one row per condition (avoids duplicates)
• question is the market title/question text
• active=true means the market is currently tradeable
• closed=true means the market has stopped trading (may or may not be resolved)
• Tags is an array — use hasAny(tags, [‘politics’]) for tag filtering
• condition_id is the primary join key to trades, positions, volume, and OI tables

• Metadata reflects current state — does not track historical changes to descriptions or end dates
• category/subcategory not directly in this table — join with curated ranking or category tables

agent.polymarket_events

Polymarket event-level metadata. Events group related markets (conditions) under a single theme (e.g., “2026 US Presidential Election” groups multiple candidate-specific markets). Useful for navigating the market hierarchy.

• Events contain multiple conditions — join with market_details on event_slug for full expansion
• Useful for building hierarchical market browsers (event -> conditions -> trades)

• Not all markets belong to events — some standalone markets have no parent event

agent.polymarket_prices_daily

Daily outcome token prices for Polymarket markets. Tracks the probability (price) of each outcome over time. Useful for building price charts and tracking market sentiment evolution.

• Price represents implied probability (0-1 range, where 1.0 = 100% chance)
• Combine with market_details for market question context
• For binary markets, only one outcome price is needed (the other is 1 - price)

• Daily granularity only — no intraday price snapshots available in this table
• Prices reflect last trade of day, not volume-weighted average

agent.polymarket_prices_hourly

Hourly CLOB mid-prices per Polymarket outcome token. Provides intraday price (implied probability) resolution for building higher-frequency charts and detecting rapid sentiment shifts.

• Prices represent implied probability (0-1 range). Multiply by 100 for percentage.
• Filter on block_date first for partition pruning, then narrow by hour if needed
• For binary markets, one outcome price is sufficient — the other is 1 - price
• JOIN with polymarket_market_details ON condition_id for market question context

• Mid-prices from CLOB — may be stale for illiquid markets with wide spreads
• Not volume-weighted — reflects order book mid, not trade-weighted average

agent.polymarket_prices_latest

Latest snapshot price per Polymarket outcome token. A convenience view that returns the most recent price for each condition, eliminating the need for argMax() queries against the full prices table.

Engine: MergeTree | ORDER BY: token_id

Column	Type	Description
token_id	String
condition_id	String	Polymarket condition ID
question	String
outcome_index	UInt8
outcome_label	String
last_updated	Date
latest_price	Float64

• One row per condition — no time filter needed
• Equivalent to argMax(price, timestamp) from the full prices table but pre-computed
• JOIN with polymarket_market_details ON condition_id for enrichment

• Point-in-time snapshot — reflects the most recent pipeline run, not tick-level
• Stale for markets with no recent trading activity

agent.polymarket_volume_daily

Daily aggregated trading volume per Polymarket market condition. Pre-computed from individual trades for efficient time-series queries. One row per (condition_id, block_date).

Engine: MergeTree | Partition: toYYYYMM(block_date) | ORDER BY: condition_id, block_date

Column	Type	Description
block_date	Date	Date of the volume observation.
condition_id	String	Market condition identifier.
trade_count	UInt64
notional_volume_usd	Float64	Daily notional trading volume (one-sided, USD)

• Much faster than aggregating polymarket_trades for daily volume — prefer this table
• JOIN with market_details ON condition_id for market context
• For platform-level total volume, sum across all conditions per day

• Granularity is daily only — for intraday volume, use polymarket_trades directly

agent.polymarket_volume_hourly

Hourly trading volume per Polymarket condition. Pre-aggregated from individual trades for efficient intraday volume analysis without scanning the full trades table.

Engine: MergeTree | Partition: toYYYYMM(toDate(block_hour)) | ORDER BY: condition_id, block_hour

Column	Type	Description
block_hour	DateTime
condition_id	String	Polymarket condition ID
trade_count	UInt64
notional_volume_usd	Float64	Daily notional trading volume (one-sided, USD)

• Filter on block_date for partition pruning, then refine by hour
• Much faster than aggregating polymarket_trades hourly
• JOIN with polymarket_market_details ON condition_id for market context

• Hourly granularity only — for sub-hour analysis, use polymarket_trades

agent.polymarket_rolling_metrics

Pre-computed rolling window metrics for Polymarket markets (e.g., 7-day volume, 30-day volume, trailing unique traders). Useful for ranking and screening markets by recent activity without expensive window function queries.

• Pre-computed rolling windows are much faster than manual window functions on trades
• Use for market ranking and filtering — then drill into volume_daily or trades for details

• Rolling window definitions are fixed — cannot customize the window period
• Updated periodically, not real-time

agent.polymarket_open_interest_daily

Cumulative open interest per Polymarket condition derived from on-chain events (splits, merges, redemptions, conversions). SPARSE: only has rows on days with events, not every calendar day. Requires gap-fill logic for time series.

Engine: MergeTree | Partition: toYYYYMM(block_date) | ORDER BY: condition_id, block_date

Column	Type	Description
block_date	Date	Date of the OI change event. SPARSE — not every day has a row.
condition_id	String	Market condition identifier.
daily_net_change_raw	Int256
daily_net_change_usd	Float64
open_interest_raw	Int256
open_interest_usd	Float64	Cumulative OI in USD for this condition on this date. Can be negative for neg-risk conditions.

• SPARSE data — use argMax(open_interest_usd, block_date) per condition to gap-fill (carry forward)
• Do NOT use HAVING last_oi > 0 — some conditions legitimately have negative OI (neg-risk markets)
• Do NOT filter by market resolution status — resolved markets’ OI decreases naturally via PayoutRedemption
• For global OI: sum(argMax(open_interest_usd, block_date)) across all conditions
• Neg-risk markets: individual conditions can have large negative OI — the group total across sibling conditions is correct
• Do NOT apply neg_risk correction (sum/count(DISTINCT condition_id)) — the raw sum is already correct

• Sparse — only rows on days with OI-affecting events, not a continuous daily series
• ~5% undercount vs Polymarket subgraph due to ~97K conditions missing from decoded events
• Individual neg-risk condition OI can be large negative — always sum across sibling conditions for market-level OI

agent.polymarket_user_activity

Per-user trade activity on Polymarket including buy/sell counts, volume, and realized PnL. Aggregated by (address, condition_id) to show each user’s activity within each market.

Engine: MergeTree | Partition: toYYYYMM(block_date) | ORDER BY: account, token_id, block_time

Column	Type	Description
block_date	Date	UTC block date
block_time	DateTime64(3)	UTC block timestamp
block_number	UInt64	Block number
tx_hash	String	Transaction hash
evt_index	UInt32	Event log index in transaction
account	String
token_id	String
event_type	String
side	String
amount	Float64	Amount in display units (decimal-adjusted)
price	Float64	Token price in USD
amount_usd	Float64	USD value of transfer

• Filter by address for single-wallet analysis
• account = maker follows Paradigm methodology — captures both sides correctly
• JOIN with polymarket_market_details ON condition_id for market context
• Aggregated per (address, condition_id) — sum across conditions for total user activity

• Uses proxy wallet addresses — no mapping to user identities
• PnL is realized only — does not include unrealized gains on open positions

agent.kalshi_market_details

Kalshi market metadata and snapshot data. Contains market tickers, titles, categories, settlement details, and status. Kalshi uses a three-level hierarchy: series -> events -> markets. MVE (parlay) markets skip the event layer.

Engine: MergeTree | ORDER BY: market_ticker

Column	Type	Description
market_ticker	String	Unique market identifier (Kalshi)
event_ticker	String	Parent event ticker. Empty for MVE (parlay) markets.
event_title	String	Event title
event_subtitle	String
title	String	Market title/question text.
start_time	DateTime64(3)
end_time	DateTime64(3)
close_time	DateTime64(3)
category	String	Market category
subcategory	String	Market subcategory
payout_type	String
status	String	Market status: ‘active’, ‘closed’, ‘settled’, etc.
last_price	Float64
total_volume	UInt64	Cumulative total volume
daily_volume	UInt64	Daily contract volume
open_interest	Float64	End-of-day open interest (contracts)
result	String	Market result (if resolved)
yes_bid	Float64
yes_ask	Float64
no_bid	Float64
no_ask	Float64
last_day_open_interest	Float64

• Kalshi hierarchy: series -> events -> markets. MVE markets skip the event layer
• event_ticker links to events; series_ticker from events links to series
• For MVE markets, derive series_ticker from splitByChar(’-’, ticker)[1]
• Use FINAL if the underlying table is ReplacingMergeTree

• Large table (55.6M rows) due to historical snapshots — filter by status or date range
• MVE (parlay) markets have different structure than regular markets — see Kalshi data model docs
• Centralized exchange data — no on-chain verification possible

agent.kalshi_trades

Individual Kalshi trade events. Each row represents a single fill on the Kalshi exchange with price, quantity, side, and market ticker. The raw trade-level data source for Kalshi volume analysis.

Engine: ReplacingMergeTree | Partition: toYYYYMM(trade_date) | ORDER BY: ticker, created_time, trade_id

Column	Type	Description
trade_id	String
ticker	String
num_contracts	UInt32
taker_side	String
maker_side	String
yes_price	Float64
no_price	Float64
trade_date	Date
created_time	DateTime64(3)

• Filter on date or created_time for partition pruning
• ticker links to kalshi_market_details for market metadata
• Price is in cents (0-100 range for binary markets)
• count is the number of contracts in the fill

• Centralized exchange data — no on-chain verification
• Trade history depth depends on Kalshi API availability
• No user/wallet attribution — trades are anonymous

agent.prediction_markets_daily

Unified daily prediction market volume and open interest across Kalshi and Polymarket. One row per (date, source, category, subcategory). Uses Paradigm-compatible category taxonomy. Only includes dates where BOTH platforms have data to avoid partial comparisons.

• source is ‘Kalshi’ or ‘Polymarket’ — filter for single-platform or compare both
• Both platforms must have data for a date to appear — gaps mean one platform had no data
• notional_volume_usd is one-sided volume in USD
• open_interest_usd is cumulative OI in USD
• For total market volume: sum both sources

• Only dates with data from BOTH platforms — single-platform activity on gap days is lost
• Category taxonomy depends on upstream categorization models — some markets may be ‘UNKNOWN’
• Volume methodology may differ slightly between platforms (both aim for one-sided)

agent.prediction_markets_matched_markets

Polymarket-to-Kalshi matched market pairs. Each row is a strict one-to-one (condition_id, market_ticker) pair linking the same event across platforms. Enriched with live volume and OI from both sides. Generated via LLM matching (Grok) with confidence scoring.

• match_type is ‘exact’ (same question) or ‘related’ (same event, different angle)
• confidence ranges from 75-100 — higher is more reliable
• One-to-one mapping — no fan-out or duplication
• JOIN with platform-specific tables for additional enrichment

• Coverage depends on LLM matching quality — not all markets are matched
• Confidence threshold at 75+ — some lower-confidence matches may be excluded
• Seed-based — new markets require re-running the matching pipeline

agent.prediction_markets_matched_daily

Daily volume and open interest comparison for matched Polymarket-Kalshi market pairs. One row per (condition_id, market_ticker, date) with side-by-side volume and OI from both platforms. The time-series companion to matched_markets.

• Filter on date for partition pruning
• Compare polymarket_volume_usd vs kalshi_volume_usd for platform share on matched events
• JOIN with prediction_markets_matched_markets for match metadata (confidence, match_type)

• Only includes matched markets — unmatched markets on either platform are excluded
• Date coverage limited to overlap period where both platforms have data

agent.polymarket_positions

Current Polymarket user positions with market context. A custom view that JOINs user_positions_v2 (on-chain position tracking) with market_details for enrichment. Shows open positions with balance, PnL, cost basis, and market metadata. Filters to positions with net_shares > 0.000001 (non-dust).

• Filter by address for wallet-specific portfolio view
• holdings_usd = net_shares * value_per_share (current market value of position)
• pnl is the realized + unrealized PnL for this position
• cost_basis is the total USD spent acquiring the position
• avg_buy_price is the volume-weighted average entry price
• resolution_status shows if the market has resolved (won/lost/open)

• Current state snapshot — does not track historical position changes
• Only shows positions with net_shares > 0.000001 — fully closed positions are excluded
• Proxy wallet address, not the user’s main wallet — address mapping not included