The same election night headline hits Kalshi, Polymarket, PredictIt, offshore books, and crypto Twitter at once—but prices rarely match for long. Cross-market arbitrage is the business of exploiting those gaps while surviving different rules, capital rails, and user pools.
This chapter is operational: how professionals think about multi-venue games, not a platform advertisement.
Why venues disagree (even with same headline)
A 6-point gap is often “different products” not “free money.”
Venue archetypes (2024–2026 landscape)
US-regulated event contract (DCM-style)
- KYC, USD banking, surveillance
- Often tight on headline US politics, expanding sports/macro
- Resolution tied to exchange rulebook + official sources
Academic / capped political (PredictIt-style)
- Low max stakes → prices can diverge because whales cannot arb
- Useful poll-like signal, bad for size
Crypto global AMM/CLOB hybrid
- USDC, wallet custody, oracle settlement
- Deep on mega headlines during peak news; geofenced US users
Traditional wagering / exchange (Betfair-style)
- Sports-first; politics where legal
- Odds format differs (decimal); commission model
OTC / institutional desks
- Block trades not visible on retail screens
Always map product not logo.
The arb matrix
Build a table for each event:
Trade only if one-liners match or you model basis risk explicitly.
Long YES cheap / short YES expensive
Canonical structure (conceptual):
- Buy YES where mid 55%
- Sell YES where mid 63% (or synthetically buy NO at 37% if complete market)
Locked profit per share ≈ 8¢ minus fees if both settle identically.
Risks:
- Leg risk — one venue halts, delists, or disputes oracle
- Margin — short may require collateral
- Currency — USD vs USDC depeg stress
Capital routing game
Arb firms maintain:
- Float on multiple venues pre-funded
- Treasury for bridge latency
- Legal entity map for who may trade where
Retail attempting arb with one bank account and MetaMask often loses to withdrawal clock—gap closed before USDC arrives.
PredictIt divergence lesson
When caps limit orders to $850 (historical magnitudes), a whale cannot push PredictIt to “true” price—but can move Kalshi. PredictIt may sit at 70% while Kalshi 62% persistently.
Interpretation: PredictIt is biased estimator (participant mix + caps), not lagging Kalshi. Media citing only PredictIt misread the game.
Polymarket vs Kalshi narrative
During peak US election liquidity:
- Often tight alignment on winner-take-all after fees
- Divergence spikes on obscure props (cabinet pick #4)
- Oracle delay on crypto side creates temporary windows
Bots monitor APIs; human edge is judging resolution diffs, not clicking obvious 10% on CNN screenshot.
Execution stack (professional)
- Normalizer — map events across venue metadata
- Fair value — weighted composite mid (liquidity weights)
- Trigger — gap > threshold + min depth
- Router — simultaneous legs, cancel if partial fill
- Hedge — options, correlated state markets
- P&L attribution — fees, failed legs, basis
Retail without stack should assume observation role, not arb role.
Regulatory game
US person on geoblocked crypto venue = ToS violation + legal exposure. Firm with licenses trades permitted pairs only.
Regulatory arb ≠ price arb: choosing legal venue is compliance, not edge.
When to cite which price
Persistent gaps as information
Sometimes gap is signal:
- Crypto venue prices regulatory approval risk higher—US venue cannot.
- Offshore prices turnout model crypto traders favor.
Ask who is in the pool before forcing convergence bet.
What comes next
Next: Cross-contract arbitrage within one platform.