Four ways prediction-market arbitrage numbers lie
An in-depth look at audit lessons from Polymarket and Kalshi order books that reveal why theoretical arbitrage edges vanish.
Stock photo for illustration only, not from the actual event
- Polymarket's Ask sort order can trick scripts into reading the worst price instead of the best.
- Execution pricing requires walking the order book ladder rather than just looking at top-of-book.
- Fee structures scaling with variance terms can entirely consume gross spreads.
- Fuzzy string matching across venues often pairs directional bets instead of true arbitrage.
After spending several months capturing and auditing order-book and oracle data from prediction markets like Polymarket and Kalshi for independent research, a developer discovered that calculated cross-venue arbitrage figures were fundamentally incorrect due to four boring yet powerful reasons that generate confident, completely useless results.
The first pitfall stems from Polymarket's CLOB returning asks in descending price order. Reading asks[0] expecting the best price inadvertently retrieves the most expensive one in the book. This issue is particularly insidious because it fails silently in a plausible direction, causing scanners to misinterpret markets as having zero liquidity and reporting no arbitrage opportunities for days simply due to a sort order issue.
Stock photo for illustration only, not from the actual event
The second error involves execution pricing assumptions. Nearly every spread screenshot compares top-of-book prices which typically cover only a handful of shares. For a 100-dollar stake, the relevant metric is the average price after walking the ladder. On thin markets, the discrepancy between top-of-book and an actual fill routinely outweighs the entire claimed edge.
The third surprise involves trading fees that many public tools incorrectly assume to be zero. Both venues charge takers a fee shaped like a variance term where Polymarket rates vary by market category, highest in crypto and lowest in politics, while Kalshi applies a similar formula rounded up to the next cent. Honest fee accounting quickly turns several perceived arbitrage opportunities solidly negative.
Understanding the structural mechanics of prediction markets clarifies why true risk-free arbitrage is exceptionally rare. Single market makers typically quote both sides of binary options tightly apart, pricing entire ladders from internally consistent models. Consequently, a thin market does not inherently imply an inconsistent or mispriced market.
The final and most instructive error relates to cross-venue market matching. Relying solely on fuzzy string similarity to pair markets across platforms will mistakenly couple directional bets disguised as arbitrage. Robust matching requires stringent hard gates covering resolution windows, numeric strikes, and event stages, accompanied by a confidence score for every row.
Source: Dev.to
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment