Applied Prediction Markets

Welcome

Prediction markets are a decentralized means of crowdsourcing information by allowing people to bet against each other on future events. They produce surprisingly accurate probability estimates of uncertain events including in sporting matches, politics, epidemiology and the weather.

In the same way what people say is less important than what they do in determining their intentions, what people wager on future events is a better predictor of the likely outcome than simply asking them. If you are a believer in the efficient market hypothesis , then prediction markets are provably the best way of extracting information in the sense that market prices ultimately reflect and aggregate all available information.

Either way, prediction markets have interesting theoretical properties and have consistently been shown to outperform other means of predicting the outcome of future events such as polling and model-based forecasting.

About me

My name is Andrew, I’m a computer science graduate from Australia. I’ve been working with prediction markets on Betfair for some time, and for the past year have had the opportunity to do it full time.

In this time I’ve become well acquinated with Betfair, and the API in all of its quirks – some of which are documented but a lot of which I’ve discovered through trial and error. I’ve also worked a lot on the engineering side, thinking about and testing how best to build and apply profitable trading algorithms. I couldn’t find a lot of information about this out there, so I decided to start writing down some things I’ve learned.

I’ll be posting updates on the first of each month, more often if I have time. If you have any questions about prediction markets, Betfair, machine learning, Python, or the intersection thereof, please reach out:

Email · GitHub · LinkedIn

Papers I like