By: Edward Egros

croatia

World Cup Finale

Pasted Graphic
Just like with statistics themselves, analyzing the results of any model can be manipulated and spun to fit a narrative. On the one hand, our World Cup model was not perfect when it came to picking the result of every match correctly. Again, our ground rules were to correctly predict if the designated "home team" would win, lose or draw in the group stage and win or lose in the knockout stage. Here are our results:

Group Stage: 27/48 (21 results were one of the two other outcomes than predicted)
Round of 16: 5/8
Quarterfinals: 2/4
Semifinals: 1/2

On the other hand, most of the misclassifications were often marked as having poor odds. For instance, for the Semifinal between England and Croatia, our model gave England a 53% chance to win. The odds were small enough to suggest extra time would be a decent possibility, and in fact
that was the outcome. Also, no other models I was actively monitoring forecasted the more unbelievable results, such as Russia knocking off Spain. In general, we are pleased with our results.

On that note, here are our predictions for the final weekend of the World Cup:

Final: France defeats Croatia (78%)

3rd Place: England defeats Belgium (65%)