← All articles
Education

Cross-Market Arbitrage

Cross-market arbs exist when a sportsbook's price differs enough from a Kalshi or Polymarket price that wagering opposite sides locks in guaranteed profit. Rarer than cross-book arbs but more durable.

April 23, 2026 · 4 min readarbitrageprediction-marketsfundamentals

Arbitrage exists when the combined implied probability of all outcomes priced across two or more books is less than 100%. Wagering a calculated stake on each outcome locks in a guaranteed profit regardless of which side wins, before friction. Cross-book arbs sit between two sportsbooks. Cross-market arbs sit between a sportsbook and a prediction market — Kalshi or Polymarket.

Why cross-market arbs exist

Sportsbooks and prediction markets are structurally different products. A sportsbook quotes a price against the house: the book takes the other side and embeds a margin (vig) to cover its operating cost and information disadvantage. A prediction market is a peer-to-peer order book: traders post limit orders against each other, and the platform charges a flat or proportional fee on filled trades rather than embedding a price spread. The two pricing mechanisms produce structurally different bid-ask configurations.

When a sharp-priced sportsbook moves on news but the prediction-market order book hasn't repriced (either because resting orders are stale or because the trader population overlaps poorly with sharp sportsbook bettors), the cross-platform price gap can exceed the sum of frictions. That gap is a cross-market arb.

How BaseCase surfaces them

The arb pipeline tags any leg involving Kalshi or Polymarket as is_cross_market = true. In the Arb Finder, cross-market opportunities render in their own section above regular cross-book arbs, marked with a cyan Cross-market pill. The visual distinction is intentional: cross-market arbs are the structurally rarest opportunities BaseCase surfaces, and they are gated to Sharp tier and above. Free users see neither the cross-market section nor any cross-market entry in their locked-row teaser.

Within each arb card, the platforms involved are labeled with platform badges. Kalshi rows additionally carry a small PM tag in the brand cyan to flag the prediction-market origin of the leg.

Q score and quality scoring

Every arb in the system carries an arb_confidence_score between 0 and 1, surfaced in the UI as a Q score from 0 to 10 (the score is multiplied by 40 and rounded, then capped at 10). The composite weighs three factors:

  1. Persistence — how long the arb has remained alive relative to the sport's median arb duration. Longer-lived arbs are more executable.
  2. Profit margin width — wider margins absorb more friction and execution slippage.
  3. Pricing normality — measured as a z-score against the consensus distribution; low z-scores indicate the arb reflects a real disagreement, while high z-scores ($\geq 3\sigma$) trigger an OUTLIER flag and a yellow accent.

Q scores at or above 6 render in green; 4-5 in cyan; below 4 in muted red. A Q score is informative but not load-bearing — operators should read both the score and the underlying margin against their own friction estimate.

Friction reality

The textbook arb assumes both legs execute simultaneously at the quoted price. In practice, both assumptions are approximate. Combined Kalshi taker fees plus sportsbook vig can run 2-4% on a typical market, which means a quoted 1.5% arb can settle at a negative net return after frictions. Burgi, Deng, and Whelan (2026) estimated an average pre-fee return of roughly -20% for Kalshi takers across a broad event sample, an order-of-magnitude reminder that the platform's fee structure is not a rounding error.

BaseCase's profit margin column is the gross figure — combined inverse-implied-probability minus 1, no friction subtraction. Treat the column as the upper bound of realized profit and apply your own platform-specific friction adjustment before sizing.

Caveats

Cross-market arbs are also exposed to resolution risk that cross-book arbs are not: Kalshi and Polymarket resolve under their platform-specific contract terms, which can differ from sportsbook grading. A sportsbook moneyline grades on regulation-plus-overtime; a prediction market may grade on regulation only, or on a conditional resolution rule. Read the contract before treating a cross-market arb as truly riskless.

Cross-market arbs are guaranteed profit in the limit of zero friction and identical resolution. Both assumptions deserve a closer look than they usually get.

Further reading