Cross-Venue Arbitrage, Measured: August 9 – August 13, 2026
Every number below is measured by our scanner from live order books and published lines over 4 days (August 9 – August 13, 2026), net of every venue fee at sizes the books could actually fill. No estimates, no extrapolation — this is the archive, summarized.
The headline numbers
- 301 distinct opportunity windows opened across 36 markets — about 75 per day. (A window is a run of detections on one market with no 30-minute silence; while these stayed open our engine logged 53,252 individual detections, one per price tick.)
- 53,203 of those detections came from prediction-market pairs (Kalshi ↔ Polymarket); 49 from sportsbook ↔ prediction-market lines.
- We tracked 329 confirmed cross-venue market pairs in this window.
- Average sportsbook overround (vig) across the 20 books we scan: 5.11% per two-way market — the baseline cost any cross-book price has to beat. Live per-book numbers: venue comparison.
How big do opportunities get?
Distribution of net edge across every logged detection:
| Net edge (after all fees) | Value |
|---|---|
| Median | 0.74% |
| Average | 1.04% |
| 90th percentile | 2.31% |
| Largest observed | 12.49% |
The distribution is heavily right-skewed: most opportunities are small, and the occasional large one appears when a single venue's price moves fast on news and the others lag by minutes.
The biggest opportunity of the window
Baltimore Orioles @ Minnesota Twins (2026-08-12 17:40 UTC) — a 12.49% net edge detected on August 12 at 14:06 UTC, with $54 of measured room at a $433 reference stake before the books converged.
How long do they last?
Of the 52,791 detections our engine re-checked 60 seconds later, 100% still cleared all fees — the rest had already been taken or moved.
Speed is the whole game: an opportunity is a live quote, not a standing offer.
Methodology
Our scanner watches Kalshi and Polymarket order books continuously and 20+ sportsbook lines every 15 minutes (pre-game only). For every candidate pair it walks both books level by level to a fillable size, subtracts every venue fee from the published schedules, and records a detection only when the net figure is positive. Each detection is re-checked seconds later and the result stored — including the ones that vanished. The raw per-market history behind these numbers is browsable on every market page.
FAQ
Can I still act on these? These are historical measurements — each one was a live quote that has since converged. The dashboard shows current ones.
Why do opportunities exist at all? Venues price independently and their customers differ. Arbitrageurs closing these differences are the mechanism that keeps venues aligned — see how it works.
Is this a promise of returns? No. It is a record of observed pricing. Execution, fills, fees, account limits, and settlement risk are all yours.
ArbMonster is a data service. Nothing here is financial or betting advice; markets carry risk, venues have age and jurisdiction restrictions, and you are responsible for verifying everything before acting.