Friday, June 15, 2018

World Cup!

The World Cup is back, and so is my World Cup model. I very much enjoyed building and publishing my 2014 World Cup model. Watching all the games, updating the model, and getting into the rhythm and sweat of each game and the tournament as a whole was just super fun. And most importantly readership was great too.

It's 2018, and time to do it all over again.

The new model is fairly similar to the 2014 version. That version performed well (notably - identifying Brazil as maybe not an overwhelming favorite), and the underlying logic was good. For this model, I start with country-level team soccer ratings, then optimize with as much available betting data as possible.

As games get played, I'll post pre-game heat maps that indicate which outcomes are most likely for that game (see more about those here). I'll also post in-game annotated graphs that give updated win probability - read more about that here!

I've added a few new things as well :)

On the right hand side of the page, I've got some detailed data about each day's game.

  • Teams playing (duh)
  • Each team's odds of winning (note: in soccer, ties are possible)
  • How each team's chance to advance from pool play depends on the game. For example, if Uruguay wins today, they'll have a 96% chance of advancing from pool play, but if they lose, their chances will be just 44%!
  • Game Scores
    • Fans of my college football work know that I think it's interesting to try to estimate which games are most interesting & worth watching. I've extended that to the World Cup. These are a work in progress so feedback is appreciated.
    • Quality score: Measures how good the two teams are. All things being equal it's more fun to watch Spain vs. Portugal than Saudi Arabia vs. Iran. A higher score means higher quality teams.
    • Competitiveness score: How close the game is likely to be. It's more fun to watch two average teams play a tight game than watch Germany beat up on some poor team. A higher score means the game is more likely to be competitive.
    • Impact score: How much impact the game has on each team's chances to advance from pool play to the knockout stage. Higher means the game is a bigger deal. For example, Spain and Portugal are likely to advance regardless of their result today, so that game has a low impact score. But Uruguay vs. Egypt is close to do or die for both teams!
On the left hand side of the page, I have each country's odds of advancing from pool play into the knockout stage. I plan to update that at the end of each workday.

Without further ado, happy World Cup!

No comments:

Post a Comment