Prediction Market Arbitrage Radar (Kalshi ↔ Polymarket)

Scan Kalshi and Polymarket simultaneously and report cross-venue matched pairs with an actionable, fee-inclusive edge — or feed it your own known pairs via manualPairs and let it price them for you every run.

What you get

One dataset record per matched pair:

• edgeCents — riskless lock-in edge in cents after Kalshi taker fees (0.07 × P × (1−P) per side). Polymarket charges no taker fee, so it is not added.
• strategy — which legs to buy: PM-YES/K-NO (Polymarket YES + Kalshi NO) or K-YES/PM-NO (Kalshi YES + Polymarket NO), computed from live asks, not last prices.
• Full quotes on both sides: yes bid/ask, Kalshi no-ask, spreads, 24h volumes, end times, URLs.
• similarity + matchMethod (fuzzy, h2h, manual) so you can judge match quality.
• riskFlags — honest warnings: end-date-gap-Nd (resolution dates differ), wide-*-spread, predicate-mismatch-style guards already exclude false pairs upstream.

Two modes

1. scan (default) — fuzzy-matches open binary markets across venues with strict guards (predicate signature, year consistency, exact team-token match for head-to-head markets). Filter with minEdgeCents, minSimilarity, minVolume24hUsd.
2. watchlist — supply exact [{"kalshiTicker":"KX...","polymarketSlug":"will-..."}] pairs; the Actor re-prices them every run. This is the reliable way to monitor markets you already know are the same event.

Why a radar, not a geyser

True cross-venue edges are transient — they appear when news hits one venue first and close within minutes. Today's scan may return one +0.4c pair or none; the value is running this on a schedule (or after news) and catching the moment. Zero-result runs are honest runs, and cost you only the run price. If you need guaranteed hits, use watchlist with pairs you track yourself.

Fees & assumptions

• Kalshi fee model: 0.07 × P × (1−P) per contract, applied to each Kalshi leg. Polymarket: 0.
• Edges use displayed asks; real fill depends on book depth at execution time.
• End-date gaps > 7 days are flagged: same question, different resolution dates ⇒ not a true arb.

Use cases

• Automated news-driven edge alerts (schedule the Actor hourly/daily).
• Monitoring your own Kalshi↔Polymarket pair book via watchlist.
• Research datasets of cross-venue spreads over time.

Input

Field	Type	Default	Description
mode	string	scan	scan or watchlist
minEdgeCents	number	0	Report threshold (fee-inclusive edge in cents)
minSimilarity	number	0.55	Fuzzy match threshold in scan mode
minVolume24hUsd	number	200	Liquidity filter per venue
maxKalshiEvents	integer	1500	Kalshi events to paginate
maxPolymarketMarkets	integer	300	PM pool size (top by 24h volume)
maxResults	integer	100	Cap on pair reports
manualPairs	array	[]	Exact pairs for watchlist mode

Output

Dataset of pair reports, sorted by edge descending. See the Overview table view in the Console, or consume the JSON directly (edgeCents, strategy, quotes, riskFlags, URLs). MCP/agent-ready: machine-readable fields, no scraping artifacts.

Sources

Official public endpoints only: trade-api/v2 (Kalshi), gamma-api.polymarket.com. No login, no cookies, no proxies, no browser.