0

I am analyzing League of Legends championships data, I am interested in the Win/Loss binary variable. The other variable would obviously be the team. Could anybody give me an idea how to do this or what my prior should be?

I wanted to use a power rankings from before the championships to have an idea of the prior for my model. This whole thing is very different than frequentist statistics and I am so lost. What kind of model should I do here I just have a binary result for each result of the Quarter Finals-Finals.

tedescr
  • 53
  • 6
  • 2
    Welcome to SO! This question is not framed well for StackOverflow at the moment. I suspect it is more of a theoretical question, in which case it is more suitable for [CrossValidated](https://stats.stackexchange.com/) (a Stack site about stats). If not, though, then please make it *reproducible*. This includes sample code (including listing non-base R packages), sample *unambiguous* data (e.g., `dput(head(x))` or `data.frame(x=...,y=...)`), and expected output. Refs: https://stackoverflow.com/questions/5963269, https://stackoverflow.com/help/mcve, and https://stackoverflow.com/tags/r/info. – r2evans Dec 18 '19 at 02:17
  • 1
    With just scores, the typical thing to do is fit a Bradley-Terry model, which is just a logistic regression with indicators for teams with Pr[team i beats team j] = inv_logit(alpha[i] - alpha[j]). If you're willing to consider score, you might want to check out Andrew Gelman's analysis in the blog post [Stan goes to the World Cup](https://statmodeling.stat.columbia.edu/2018/06/15/stan-goes-world-cup/) or Milad Kharratzadeh's StanCon case study [hierarchical modeling of the Premiere League](https://github.com/stan-dev/stancon_talks/tree/master/2017/Contributed-Talks/02_kharratzadeh) – Bob Carpenter Dec 18 '19 at 18:05
  • If you are really interested in just priors, then there is a wiki that the Stan developers setup with some advice: https://github.com/stan-dev/stan/wiki/Prior-Choice-Recommendations – Kreuni Dec 23 '19 at 02:31

0 Answers0