Polymarket × Kalshi Arbitrage Scanner

Tier 1 vs Tier 2 vs Tier 3 — what each gate means

Tier 1 is the tightest band. To clear it, an opportunity has to be cross-platform (Polymarket × Kalshi), survive fees and depth, have fresh quotes on both legs, and pass our weakest-link check. Net edge has to be positive after the conservative fee model — not gross. We hit Tier 1 rarely. That's the right shape: if the gate were loose, Tier 1 would be noise.

Tier 2 covers everything that is structurally a mispricing but doesn't survive Tier 1's gates — usually because spread eats the edge, depth is thin, or one leg has stale quotes. Worth watching, not worth firing.

Tier 3 surfaces single-platform inconsistencies (Polymarket sum violations, Kalshi monotonicity breaks) and low-volume or stale signals. Useful as a market-quality canary even if you don't trade them.

How fees and net edge are modeled

Gross edge is what a naive arbitrage calculator would show. Net edge is what's left after you actually trade. We model:

• Polymarket: dynamic taker fee scaled by the platform's posted schedule.
• Kalshi: 0.07 · C · P · (1 − P) per Kalshi's published formula.
• Spread: 1–3% on each leg, depending on order-book depth at the size you're sizing for.
• Capital cost: a small annualized drag on locked capital while the position settles.

Tier 1 requires net edge to clear a positive threshold after all four. Tier 2 includes opportunities where gross edge looks fine but net edge is fee-tight. The gross/net split is shown on every card so you can see the haircut, not just the headline number. Once you've picked a Tier 1 to hit, size the position with the Polymarket Kelly calculator — it pulls the live opportunity data and applies credible-bound shrinkage.

Detection types — sum violation, monotonicity, cross-platform

The scanner runs six detection passes against canonical-key matched events between Polymarket and Kalshi.

• cross-platform: same outcome priced differently on the two platforms. Buy the cheaper YES, buy the matching NO on the other side. The classic.
• cross-platform bracket: outcomes bracket each other rather than match exactly (e.g. Poly ≥ X, Kalshi ≥ Y). Sized via the worst-case overlap.
• polymarket sum violation: across a multi-outcome Polymarket market, YES probabilities sum to less than 1 (or NOs sum to more than 1). Buy all YESes; profit at resolution.
• polymarket monotonicity: ordered Poly markets where a strictly-broader outcome trades cheaper than a stricter one.
• kalshi monotonicity: same pattern on Kalshi tier markets.

Every detection is logged with the underlying event, leg prices, and the canonical key used to match. The full methodology page covers the gate thresholds, fee-model derivation, and how canonical keys are constructed for cross-platform matching.

Weakest-link transparency — and why it's the rule

Every Tier 1 card carries a weakest_link_summary — the single weakest input to the trade. It might be "Polymarket leg has a 12-second stale quote", "Kalshi depth caps you at $340 net edge", or "implied YES on Poly equals 1 − NO sum but resolution language differs by 2 days".

If the scanner can't write a weakest-link summary, the opportunity does not promote to Tier 1. That's the rule, not a guideline. The weakest link is the actual edge you're betting on — surfacing it lets you decide if you're comfortable owning it. We'd rather show fewer Tier 1s than show a confident card whose weakest link was a stale quote we hid.