Sports We Cover
Markets
Our Data
How it works
Scraping
We scrape nightly over several important sources of data for every sport that we predict for: (1) base statistics and reference data about teams and players, (2) auxilliary data, such as expert picks, social media, and latest news, and (3) historical betting odds and odds trends.
Machine Learning
We clean our scraped data transform our raw data into meaningful signals through feature engineering. In general, for each sport we will have between 100 – 1000 predictive signals. We then train machine-learning logistic models for winning probabilities. The majority of our models boosted gradient decision trees, but occasionally we will also use deep neural networks.
Backtest Strategies
Betting on individual events yields a binary payoff with high variance. Also, some events occur simultaneously and cannot be bet in sequence. We provide a suite of betting strategies to determine how much capital to allocate to each sporting event. Our primary strategy adjusts variance-adjusted returns (maximizes Sharpe ratio), but we also offer strategies that chase higher returns, or minimize variance.
Subscription Service
We offer a subscription service to our data, via REST endpoints or via transfer of csv files. Please contact us for a free consultation on how our rates work and for the pricing tiers for our subscription data.
View the performance of our models
Who We Are
Our Team
Oliver Huang
Data ScientistVaughn Lorie
OperationsGet In Touch
Contact
Please reach out to us for details on our subscription packages, or for questions and suggestions about our project.
We've collaborated with:
- amateur betters who want to beat the market
- sports analytics professionals who want to conduct research using our data
- bookmakers who want to use our data for odds market-making
- sports data vendors
To get in touch, fill out the form on the left, or feel free to email us directly.