The most recent Elo World Football Ratings were published on 3 January 2019 (link) and the FIFA Rankings on 20 December 2018 (link).
At the start of the 2019 Asian Cup (link), the ratings and rankings of the teams in the competition are:
In the 2015 Asian Cup tournament, the performances against Elo ratings were:
- Higher Elo rated team scores first and wins: 22 games (69% of games played)
- Higher Elo rated team does not score first and wins: 2 games (6%)
- Higher Elo rated team scores first and loses: 3 games (9%)
- Higher Elo rated team does not score first and loses: 5 games (16%)
I am going to monitor performance against Elo rating in 2019 and will start with these naive priors:
- 0.69 Higher Elo rated team scores first and wins
- 0.75 Higher Elo rated team wins
- 0.25 Higher Elo rated team loses
The tournament starts this evening with the United Arab Emirates v Bahrain Group A game in Abu Dhabi (link).
Hi Keith – as always, great post. Elo is a terrific method but regarding the prior, “the higher rated Elo team” will always have a higher probability of winning (choosing .75 will correspond to an Elo rating difference of somewhere around 200 pts). However, the difference between Elo ratings is what dictates how much more likely the higher rated team is to win. When asked “how well does Elo predict outcomes?”, the correct answer is always correct “it depends”. For teams with relatively similar Elo ratings, the probability the (modestly) higher rated will win will be just over 50%. For teams with highly disparate ratings, the win expectancy can be quite high. I look forward to your updates!
Hello, Peter. Thank you for finding the post and for your excellent points. My priors are naive and I am keen to see how they turn out. As I was writing the post a part of James Somers’ article in the New Yorker was nudging me, namely, “David Silver, the head of research at DeepMind, has pointed out a seeming paradox at the heart of his company’s recent work with games: the simpler its programs got—from AlphaGo to AlphaGo Zero to AlphaZero—the better they performed. “Maybe one of the principles that we’re after,” he said, in a talk in December of 2017, “is this idea that by doing less, by removing complexity from the algorithm, it enables us to become more general.” https://www.newyorker.com/science/elements/how-the-artificial-intelligence-program-alphazero-mastered-its-games I do hope this reply finds you well and looking forward to an exciting 2019.