>

Blog

Writing on quantitative finance, AI, and whatever else I'm building.

IDPOSTDATE
001

Why Prediction Markets are Cool

In 1591 the Pope made betting on his own succession punishable by excommunication, and the objection was never that the odds were wrong. Five hundred years of people trying to unplug the only instrument that pays for the truth.

Prediction MarketsFinanceMarketsJUL 2026
002

The Fight Futarchy Doesn't Solve

Futarchy's market half is genuinely honest; the dishonesty just moves one step earlier, into whoever gets to define the number it's betting on.

Prediction MarketsMarket DesignGovernanceJUL 2026
003

Markets Don't Beat Polls, They Absorb Them

Prediction markets look like they beat polls because they're partly made of polls, updated in real time, so the right test isn't who wins, it's what breaks when you remove either one.

Prediction MarketsForecastingElectionsJUL 2026
004

Kalshi Won on Jurisdiction, Not on the Merits

Kalshi's nationwide election markets exist because a judge decided the CFTC overstepped its statutory lane; nobody in a robe has ever ruled that betting on elections is good policy.

Prediction MarketsRegulationMarket DesignJUL 2026
005

The Bid-Ask Spread Is Three Different Prices

A two-cent spread and a twenty-cent spread aren't different amounts of the same thing; they're answers to three unrelated questions, and collapsing them into one "liquidity" number is how you mistake a capital problem for an information problem.

Market MicrostructureTradingLiquidityJUL 2026
006

The Latency Arms Race Was a Design Bug, Not a Rivalry

The decade-long HFT speed war wasn't a natural rivalry between firms; it was the predictable output of one rule in the continuous limit order book, and the moment exchanges gave traders a way to commit to an execution time, the race quietly stopped paying.

Market MicrostructureMarket DesignTradingJUL 2026
008

Regime Change Is a Clustering Problem, Not a Feeling

If "we're in a new regime" can't be tied to a number you specified before the drawdown, it's not a market observation; it's an alibi.

QuantTradingVolatilityJUL 2026
009

Polymarket's Trade-Flow Data Is Barely Better Than a Coin Flip

Feed-inferred trade direction on Polymarket matches on-chain settlement truth about 59% of the time, worse than a market with no ground truth at all should ever be, and almost nobody selling "smart money" flow is checking it.

Prediction MarketsMarket MicrostructurePolymarketJUL 2026
010

The Cheapest Edge Is a Market Disagreeing With Itself

A prediction market can call the future wrong and still be efficient, or call it right and still be riddled with contradictions; Polymarket's own NBA order books show why the second failure is the one worth watching.

Prediction MarketsMarket MicrostructurePolymarketJUL 2026
011

The Favorite-Longshot Bias Doesn't Care How Smart You Are

Polymarket's order book (crypto-native, algo-heavy, open 24/7 to professional flow) prices longshots worse than the racetrack does, which means "dumb bettors" was never the real explanation.

Prediction MarketsQuantMarket MicrostructureJUL 2026
013

Favourite Longshot Bias

Cheap prediction-market contracts win far less often than their price says they will. 64,507 resolved contracts across Polymarket and Kalshi, and at prices you can actually trade it costs you more than half your money.

Prediction MarketsQuantOriginal ResearchJUL 2026