Polymarket Arbitrage: Kalshi vs Polymarket Price Gaps (2026)

Kalshi-vs-Polymarket arbitrage buys the same event on both venues when their prices diverge — YES on one, NO on the other, for less than $1 combined. The gaps exist for structural reasons (different user bases, different fee models, different settlement rails) and persist because closing them requires capital and access on both platforms. Persistent is not permanent: the executable edge is measured in minutes.

Why do the gaps exist at all?

What does a real arbitrage look like?

Same event, matched contracts:

Combined cost: 95¢. Whoever wins, one leg pays $1.00 — a locked 5¢ gross per matched share before fees. The work: confirm the contracts truly match (rules, dates, resolution sources), compute both venues’ fees, size to the thinner book, execute both legs near-simultaneously, and manage the residual: one leg fills, the other doesn’t — now you hold a directional position.

What does the fee math decide?

Gross spread means nothing net of fees:

• Polymarket taker fee — shares × feeRate × p × (1−p); at mid-range prices in crypto categories that’s up to $1.75 per 100 shares; politics is cheaper; geopolitics is free
• Kalshi trading fee — per-contract, shown before the trade
• Exit costs — if you can’t hold both legs to resolution, unwinding costs a spread each

A 2% gross gap with 2.5% round-trip fees is a trap, not an arbitrage. The scanner math must be fee-adjusted end to end.

What are the execution risks?

1. Leg risk — one fill, one miss: the hedge becomes a naked position
2. Rule mismatch — “same” event resolving differently across venues (different sources, edge-case wording); the locked trade isn’t locked
3. Resolution timing mismatch — one venue settles days before the other; capital is pinned on the lagging leg
4. Speed competition — professional firms run this exact loop; the gaps you see in a screenshot are already gone

What does the bot architecture look like?

1. MATCHER   — align events Kalshi ↔ Polymarket (title+rules+date hashing)
2. PRICER    — poll/stream both books; normalize to implied probabilities
3. SCANNER   — flag combined cost < $1 − fees − safety margin
4. EXECUTOR  — dual-track order placement with cancel-on-partial
5. RECONCILER— positions and fills reconciled across both venues

The matcher is the moat: naive title-matching produces false positives that cost real money at resolution. Rules text, end dates, and resolution sources belong in the matching key.

FAQ

Q: Is Kalshi vs Polymarket arbitrage still profitable in 2026? A: In pockets — the structural differences keep regenerating gaps, but the executable window is minutes and favors automated, fee-aware operations over manual trading.

Q: What is the typical gap size? A: Persistent 1-5% gross between venues on matched events; after fees and leg risk, the realistic net for a well-built bot is a fraction of that, at volume.

Q: Can I arb between Polymarket global and Polymarket US? A: Mostly no for retail — they’re separate legal entities with separate apps, and US users can’t trade the global book. The classic pair is Kalshi vs Polymarket global.

Q: What’s the biggest mistake in cross-venue arb? A: Matching on titles. Two markets can ask near-identical questions and resolve differently on edge cases — the resolution rules, not the headline, define the contract.

Q: Do I need capital on both platforms? A: Yes — and collateral on each side settles in its own rail (USD on Kalshi, pUSD on Polymarket), so rebalancing between venues has its own friction and timing.

Sources

• Claw Arbs — Kalshi vs Polymarket arbitrage
• docs.polymarket.com — Fees

Disclaimer

Not financial advice. Educational purposes only. Updated October 2026.