Market‑maker margins on Kalshi
What happened
An analysis of 72 million Kalshi trades (about $18 billion volume) found market makers extracted roughly −1.12% from retail takers on average and up to 7.32% on high‑emotion events. That framing positions some prediction markets as venues where retail faces sizable effective costs versus professional liquidity providers. Risk and strategy teams should model venue microstructure impact on execution cost and slippage for retail‑facing products. (x.com)
Why it matters
A software engineer, Jonathan Becker, published an open analysis that reconstructed Kalshi’s full trade history and ran statistics across 72.1 million trades, $18.26 billion in notional volume, and 7.68 million distinct markets; his data collection ended on November 25, 2025. (jbecker.dev) (github.com) Becker’s tables show volume is highly concentrated: sports markets account for roughly 72% of notional volume, politics about 13%, and crypto about 5%, and sports became dominant after Kalshi added sports markets in 2025. (jbecker.dev) On execution roles: a “maker” is a participant who places an order that sits on the order book waiting to be filled (a limit order), while a “taker” is the participant who accepts an existing quote to get immediate execution (a market order); Kalshi explicitly records which side initiated each trade, allowing per-role return decomposition. (karlwhelan.com) (github.com) Becker documents a classic “longshot bias” — traders overpay for very-low-probability “YES” contracts (for example, contracts priced at 5¢ historically paid out only about 4% of the time) — and shows that this behavioral skew, combined with asymmetric order flow, produces a persistent wealth transfer from takers to makers; several summaries and reproductions of his figures report an average taker shortfall of about 1.12% per trade and category gaps that exceed 7% in high-emotion markets like media and entertainment. (jbecker.dev) (odaily.news) (news.ycombinator.com) Those measured gaps are a microstructure effect, not an exchange fee: as professional market makers entered and began posting deep quotes, the equilibrium shifted from a taker-friendly regime to one where posting liquidity and capturing the spread became the higher-probability profit source, while finance-oriented contracts show near-efficiency with maker–taker differences around a few tenths of a percent. (jbecker.dev) (news.ycombinator.com) Becker’s project is reproducible: the public GitHub repository contains the Parquet trade and market dumps plus Python analysis scripts used to generate the figures, and several community forks and dashboards have already been built on the same dataset, which makes it possible to backtest role-specific execution models or replicate category-level slippage measurements. (github.com 1) (github.com 2)
Key numbers
- An analysis of 72 million Kalshi trades (about $18 billion volume) found market makers extracted roughly −1.12% from retail takers on average and up to 7.32% on high‑emotion events.
- (jbecker.dev) (github.com) Becker’s tables show volume is highly concentrated: sports markets account for roughly 72% of notional volume, politics about 13%, and crypto about 5%, and sports became dominant after Kalshi added sports markets in 2025.
Quick answers
What happened in Market‑maker margins on Kalshi?
An analysis of 72 million Kalshi trades (about $18 billion volume) found market makers extracted roughly −1.12% from retail takers on average and up to 7.32% on high‑emotion events. That framing positions some prediction markets as venues where retail faces sizable effective costs versus professional liquidity providers. Risk and strategy teams should model venue microstructure impact on execution cost and slippage for retail‑facing products. (x.com)
Why does Market‑maker margins on Kalshi matter?
A software engineer, Jonathan Becker, published an open analysis that reconstructed Kalshi’s full trade history and ran statistics across 72.1 million trades, $18.26 billion in notional volume, and 7.68 million distinct markets; his data collection ended on November 25, 2025. (jbecker.dev) (github.com) Becker’s tables show volume is highly concentrated: sports markets account for roughly 72% of notional volume, politics about 13%, and crypto about 5%, and sports became dominant after Kalshi added sports markets in 2025. (jbecker.dev) On execution roles: a “maker” is a participant who places an order that sits on the order book waiting to be filled (a limit order), while a “taker” is the participant who accepts an existing quote to get immediate execution (a market order); Kalshi explicitly records which side initiated each trade, allowing per-role return decomposition. (karlwhelan.com) (github.com) Becker documents a classic “longshot bias” — traders overpay for very-low-probability “YES” contracts (for example, contracts priced at 5¢ historically paid out only about 4% of the time) — and shows that this behavioral skew, combined with asymmetric order flow, produces a persistent wealth transfer from takers to makers; several summaries and reproductions of his figures report an average taker shortfall of about 1.12% per trade and category gaps that exceed 7% in high-emotion markets like media and entertainment. (jbecker.dev) (odaily.news) (news.ycombinator.com) Those measured gaps are a microstructure effect, not an exchange fee: as professional market makers entered and began posting deep quotes, the equilibrium shifted from a taker-friendly regime to one where posting liquidity and capturing the spread became the higher-probability profit source, while finance-oriented contracts show near-efficiency with maker–taker differences around a few tenths of a percent. (jbecker.dev) (news.ycombinator.com) Becker’s project is reproducible: the public GitHub repository contains the Parquet trade and market dumps plus Python analysis scripts used to generate the figures, and several community forks and dashboards have already been built on the same dataset, which makes it possible to backtest role-specific execution models or replicate category-level slippage measurements. (github.com 1) (github.com 2)
