The concepts of win probability in sports fascinates me. I use it frequently when trying to understand the value of players or the impact of events within a cricket match
However, win probability models in any sport can seem like a black-box - even when I use my own models, I am required to trust my previous work and take results as an article of faith. I want to make my analysis as transparent as possible to avoid the black-box nature of win probability. Greater transparency holds me to a higher level of intellectual rigour and allows other people to participate in the analysis
It took a lot more effort than I had expected to write-up my methods. During the process, I discovered several features of the models which annoyed me and I ended up re-building a lot of stuff from scratch. This made very little difference to the overall results and I do wonder whether or not it was worth the time investment!
Here is the detailed write-up of my methodology: Win Probability Model