Yes, this should also be possible when using the Poisson Binomial. You can use the approach outlined by @centralwinger.com that we refer to in the blog.

TBH, I think that would also be the best way to do it for MCMC as then you avoid using rejection sampling.

@probberechts.bsky.social has also provided an online tool that computes the exact probabilities:

dtai.cs.kuleuven.be/static/sport...

This is joint work with @janvanhaaren.be & @probberechts.bsky.social

Many others have looked at the MCMC approach (cf @danny.page's online simulator) or discussed how to use the Poisson Binomial distribution to compute these probabilities exactly (cf @tonyelhabr.bsky.social, Jonas Lindstrøm,...)

3. Discuss how to foster such collaborations. We provide (anecdotal) advice, wrt possible pitfalls that may arise and how to avoid them when collaborating. We believe joint events can help facilitate collaborations. This paper arose out of discussions Sigrid and I had at such an event. 4/4

2. Highlight emerging trends in computational research on sports both from a methodological and task perspective.

Tackling these problems requires combining sports & computational expertise. I.e., deeper collaborations between sports and computational researchers will be essential.

3 / 4

1. Overview historically prominent lines of work in computational approaches (e.g., AI/ML/Statistics) to analyzing tactics in soccer.

Hopefully, this can serve as an introduction for those wanting to learn about the area but many not know where to look for relevant publications.

2 / 4