Которая решает проблему недооценки вероятности счета 0:0 и немного по-другому предсказывает количество забитых голов.
Predicting Football Results With Statistical Modelling: Dixon-Coles and Time-Weighting
https://dashee87.github.io/football/python/predicting-football-results-with-statistical-modelling-dixon-coles-and-time-weighting/
This post describes two popular improvements to the standard Poisson model for football predictions, collectively known as the Dixon-Coles model
In an earlier post, I showed how to build a simple Poisson model to crudely predict the outcome of football (soccer) matches. In the same way teams herald slight changes to their traditional plain coloured jerseys as ground breaking (And this racing stripe here I feel is pretty sharp), I thought I’d show how that basic model could be tweaked and improved in order to achieve revolutionary status.
The changes are motivated by a combination of intuition and statistics. The Dixon-Coles model (named after the paper’s authors) corrects for the basic model’s underestimation of draws and it also incorporates a time component so that recent matches are considered more important in calculating average goals rate. This isn’t a particularly novel idea for a blog post. There are numerous implementation of the Dixon-Coles model out there. Like any somewhat niche statistical modelling exercise, however, they are mostly available in R. I strongly recommend the excellent opisthokonta blog, especially if you’re interested in more advanced models. If you’re not interested in the theory and just want to start making predictions with R, then check out the regista package on GitHub.
As always, the corresponding Jupyter notebook can be downloaded from my GitHub. I’ve also uploaded some Python files, if you’d prefer to skip the highly engaging commentary.
Комментарии
Отправить комментарий