2018 Fifa World Cup Predictions
All eyes will be on Russia for soccers coveted trophy
Unless you are a baseball fan, and even if you are a baseball fan, the middle of the summer is typically a dull time for sports fans. There is the always fun NBA Draft and summer league, story lines of off-season changes, but that is about all we can hope for. However, ever four years we get the one sporting contest we forgot we loved the most — the Fifa World Cup.
Though the U.S. is left out of the 32 nations in embarrassing fashion, there will nevertheless be a wonderful display of the beautiful game in Russia.
Modeling international soccer comes with its perils, but the elite nations typically remain at the top of the world. Our model, which you can read about the methodology below, gives the nations of Germany, Brazil, and Spain the best chances to be holding the trophy in late July — the same three that Vegas has most likely to win.
Our model is much more conservative than both Vegas and others, not giving any team more than a 10% chance to win it all from the outset. It further gives several nations, namely Iran and Australia, a greater chance of success than those who follow the game on the pitch would imagine.
The model projects several Groups with ample parity that will provide soccer (or football, if you will) fans entrenched for weeks. And as you become a fan of a country you hold no relationship to for the next few weeks, take a look at our interactive RShiny App that will update with each country’s chances of advancing to different stages of the World Cup tournament here.
Methodology
International soccer is a tricky thing to predict because teams play so sporadically throughout what is basically a four-year season. And, even when they do play, they may not always display the best lineup they would on the grandest of stages at the World Cup.
Additionally, soccer statistics going back decades are incredibly sparse — limited to only the final score. Thus, we set out to build a model, using all international games going back to the late 1800s, to predict this summer’s World Cup. The model is shaped in two facets: weighting the type of competition, such that a World Cup qualifier is more important than a simple international friendly, and weighting based on the recency of the game.
The games were weighted on a logistic curve so that more recent games would be weighted more heavily than games much further away. Then, we adjusted each nation’s offensive and defensive scoring numbers based on competition of their several hundred matches individually. In order to predict each game we utilized a Poisson Process and then used Monte Carlo simulations to simulate both the group stage and the knockout stage. We ran all of the simulations “hot”, meaning that a team’s ratings were dynamic, allowing them to go on a run (for better or for worse) during the simulated tournament.
This project was a joint effort in conjunction with Delaney Ambrosen. Please reach out to me with comments or questions ]at aepowell95@gmail.com and find the code for this project and others at his github page: https://github.com/powellae.
