This paper examines arbitrage opportunities on Polymarket, a prediction market platform built on public blockchain infrastructure. The authors identify a critical design flaw in how dependent outcomes are priced, allowing guaranteed profit through coordinated position-taking.
Abstract
While outcome prices should collectively equal $1, dependent assets are mispriced — "allowing for purchasing (or selling) a certain outcome for less than (or more than) $1, guaranteeing profit." The authors address three key research questions: what conditions enable arbitrage, whether it occurs in practice, and evidence of exploitation. Using on-chain order book data, they document approximately 40 million USD in realized profit extraction.
The Core Design Flaw
In a well-functioning prediction market, the sum of prices for all mutually exclusive outcomes of an event should equal exactly $1. When this invariant is violated — whether within a single market or across correlated markets — a risk-free profit opportunity exists. The authors demonstrate that on Polymarket, such violations are both systematic and exploitable at scale.
Two Types of Arbitrage
- Market Rebalancing Arbitrage: Within single markets — occurs when outcome prices within one market fail to sum to $1 due to liquidity imbalances or delayed price discovery.
- Combinatorial Arbitrage: Across multiple markets — exploits logical dependencies between related markets (e.g., "Candidate A wins state X" and "Candidate A wins presidency") that are priced inconsistently.
Methodology
To manage computational complexity across numerous markets, the authors employ "a heuristic-driven reduction strategy based on timeliness, topical similarity, and combinatorial relationships." This narrows the search space from the full combinatorial explosion of market pairs to tractable subsets most likely to exhibit exploitable mispricings.
The analysis uses on-chain order book data — a key advantage of blockchain-based prediction markets — providing a complete, tamper-proof record of all trades and price history, enabling precise profit attribution.
Key Findings
The researchers document approximately $40 million USD in realized profit extracted through these arbitrage strategies. This represents a substantial transfer from uninformed market participants to sophisticated arbitrageurs, raising questions about the information efficiency and fairness of decentralized prediction markets.
"Dependent assets are mispriced, allowing for purchasing a certain outcome for less than $1, guaranteeing profit — a fundamental design flaw in how Polymarket handles correlated outcomes."
Implications
The findings have direct implications for prediction market design, decentralized finance protocols, and the broader question of whether blockchain-based markets can serve as reliable information aggregation mechanisms when systematic mispricings persist at this scale.