Cross-contract fixes local mistakes. Structural arbitrage optimizes across an entire event graph: dozens of states, rounds, running mates, vote bands, and timing clauses—finding portfolios that lock profit if any coherent probability assignment exists.
This is quant territory; retail benefits from knowing it exists because bots smooth the tree you trade.
What is a probability tree?
Nodes = events; edges = conditions. Each complete path from root to leaf is a world state.
Example (simplified US election):
- Win nomination?
- Win swing state bundle?
- National margin > 3%?
- Popular vote winner?
Listed contracts are marginals of hidden joint distribution. Coherence requires existence of joint P matching all quoted marginals (within spread).
Inconsistency types
Structural arb searches all listed constraints simultaneously.
Linear programming intuition
Imagine unknown probabilities p_i over atomic states (each leaf). Each market quote imposes inequality:
Σ p_i over {B wins} = 0.55 ± εΣ p_i over {turnout > 60%} = 0.30 ± ε
If constraints are infeasible, no real distribution exists → arb portfolio constructed.
Practitioners use linear / integer programs or SAT solvers for feasibility; retail uses intuition on 3-node trees.
Dutch book connection
A Dutch book is a set of bets guaranteeing loss for someone who accepts all sides at stated prices. Incoherent tree ⇒ Dutch book against passive liquidity providers.
Arb bots are Dutch book enforcers—profit while restoring coherence.
Electoral college complexity
US presidency is not “sum of states = national.” Joint distribution has constraints (270 EV logic).
Naive arb comparing national YES to sum of state YES fails—structural models embed simulation or historical covariance.
Sophisticated shops run Monte Carlo on state prices → national fair value → trade divergence.
Time structure
Trees include dates:
- Nomination by April
- Withdrawal by June
- Election November
Monotonicity: later survival probability ≥ earlier survival for same candidate path (ceteris paribus).
Violations snap quickly on liquid platforms.
Liquidity-weighted coherence
Bots may ignore illiquid nodes to avoid moving dust markets. Structural arb on $200 contract may persist while headline coherent—do not generalize.
Platform lifecycle
- List many contracts for SEO and engagement
- Misprice correlations initially
- Arb firms connect API
- Tree tightens on majors; tails still wild
Trading day 1 of multi-contract menu = rich structural arb; day 30 on election night = crumbs.
Risk beyond math
Resolution divergence — two contracts use different data sources; “structural arb” loses at settlement.
Delisting — leg vanishes mid-trade.
Rule change — platform amends criteria mid-event.
Tools for serious traders
- Export all markets JSON for event
- Build adjacency graph (shared tags)
- Run feasibility check script
- Paper trade portfolio before size
Open-source research repos demonstrate graph arb on election data—study for patterns, not copy-paste live.
Media literacy
Headline: “Markets say 78% chance X” while sibling contract implies 65%. Structural incoherence means quote wrong market or stale leg.
Ethical frame
Arb improves coherence, not truth. Tree can be coherent and collectively wrong on fundamentals (everyone misreads polling error).
What comes next
Next: Information cascades and herd behavior.