Cross-Venue Spreads
A study of price gaps between Kalshi and Polymarket on the same events, measured after fees.
What it does
Two exchanges listing the same event should price it the same way, but they often don't. Measuring whether those gaps are real means confirming the contracts are actually equivalent and pricing the fees correctly.
Technical details
- Venues
- Kalshi and Polymarket
- Capture
- 700,912 live order-book snapshots over two days, across 14 events (MLB games, tennis matches and one Federal Reserve decision)
- Matching
- Text similarity proposes contract pairs, and I confirmed each one by hand.
- Scope
- Paper simulation. No real money and no orders placed.
Results
Fee model
I built the fee model from 12,181 real fee schedules. Fees are quadratic in price and peak at 50 cents, which is exactly where most gaps occurred, so pricing them correctly decides which gaps are real.
Overall result
The average gap was +1.8% before fees and -0.25% after. 74 of 223 survived fees, averaging +1.4%. In simulation, 40 of 42 fills closed profitably, from a small two-day sample.
Speed
Gaps lasted 9.1 seconds on average, so catching them depends on execution speed more than analysis.
Contract matching
Contracts with identical wording can settle differently when the two exchanges define the outcome differently, so text similarity only proposes pairs and I confirm every one by hand.