Credits

• Henry Stott : I have and continue to learn a great deal from Henry's ingenious ways of analysing data to answer challenging questions as well as his understanding of a wide array of analytical approaches and knack for explaining how models/methodologies work in an succinct, elegant and approachable way
• Mark Latham : Mark has taught me most of the applicable statistics knowledge I have. From Markov chains to constructing and evaluating custom estimators through log-likelihood maximisation, via censored-data survival analysis, with plenty of interesting insights along the intellectually-scenic journey.
• Paul Jackson : Paul taught me a lot about software development and systems engineering. It's largely thanks to Paul that I'm now equipped with knowledge of Java and SQL. The latter has been particularly useful for keeping the underlying data for this site clean and tidy
• Kostas Paraschakis : Kostas was the first person to introduce me to GLMs, the Dixon-Coles model and showed me several predictive performance improvements he'd made to it. His aversion to methodologies that were insufficiently statistically rigorous was also very influential early on to properly assess the underlying assumptions behind models and ensure I'm using the appropriate statistical techniques and models for the problem at hand
• Daniel Finkelstein : I've been fortunate enough to have had the opportunity to work with Daniel as the main analyst behind his football analysis weekly column in the Times, the "Fink Tank" for 4 years. Working with Daniel on the Fink Tank has allowed me to analyse many interesting aspects of football and his articles have been great examples of how to communicate technical analysis results to a wider audience.

• Ben Torvaney : This project started off with me trying to create a Python analogue of Ben's great Regista package for R, which ended up sprawling into a different type of project. Ben's blog Stats and Snakeoil has also been a strong source of inspiration. The team strength estimation / match forecast model used in this site is very strongly based on Ben's approach described in this article.
• John Burn-Murdoch : JBM's steady stream of consistently excellent data visualisations for the FT are a source of inspiration on tap. They're all elegantly presented, intuitive to read, informative and interesting. The league forecasts pages on this site borrow heavily from JBM's work for the Times in 2017.
• Peter McKeever : Peter makes beautiful, slick and informative data visualisations for OptaPro. Some of the aesthetic and datavis choices made in this project have been inspired by Peter's work.
• David Sheehan : David's blog post and Jupyter notebooks explaining how the Dixon-coles model works, with working examples, were a useful starting point to develop and build my models on.
• Lars Schiefler / ClubELO : ClubELO is a football predictions site driven by the ELO rating system. I've drawn inspiration from many presentation choices made by that site, particularly around adding manager context to team strength timelines.
• Vincent D. Warmerdam : Vincent's excellent PyData talks and instructional videos on calmcode.io have helped me write cleaner code, create more elegant models, and have definitely seeded many analytics ideas. I've also learned about lots of useful Python development tools from the latter. The design of the StatsLab page in this site borrows heavily from calmcode.io

• Understat : Shots data and fixture lists
• FiveThirtyEight : European competition fixture lists and results
• API-Football : Match results and event timestamps
• TransferMarkt : Manager history
• Football-Data.co.uk : Historic fixtures and match results

This project was built entirely on an open-source software stack.

I'd like to thank the developers behind Python, Pandas, Flask, Numpy, Matplotlib, Statsmodels and Scikit-learn for creating the wonderful tools on which this site is built