Time for our last update! Qatar 2022 World Cup is progressing fast, and only four teams – Argentina, France, Croatia and Morocco – are still in contention for the final victory. Who will be the winner on December 18th? Is our model better than the Paul the Octopus, an almost perfect oracle during World Cup 2010?

Semifinals predictions

We report in the table below the posterior predictive match probabilities from our DIBP model – get a look also here and here for other updates – for the two semifinals planned for Tuesday, December 13 and Wednesday, December 14, Argentina-Croatia and France-Morocco, respectively. We also report the usual ppd ‘chessboard plots’ for the exact outcomes in gray-scale color.

Notes: ‘mlo’ in the table denotes the ‘most likely result’ , whereas darker regions in the plots correspond to more likely results. The first team listed in each sub-title is the ‘favorite’ (x-axis), whereas the second team is the ‘underdog’ (y-axis). The 2-way grid displays the 2 held-out matches in such a way that the closest match appears in the left panel of the grid, whereas the most unbalanced match (‘blowout’) appears in the right panel. 

France and Argentina seem clearly ahead against Croatia and Morocco, respectively.  Anyway, underdogs  such as Morocco have a non-negligible chance – approximately 35% – to beat France and advance to the final: consider that  Morocco got two ‘clean-sheets‘ in the round of 16 and quarter of finals matches, against Spain and Portugal, respectively!   Croatia already achieved the final four years ago, so maybe it should not be considered as a pure underdog…and Luka Modric, the Croatia’s captain, is still one of the best players in the world.

Note: keep in mind that the above predictions refer to the ‘regular’ times, not to the extra times! Anyway, to get an approximated probability to advance to the final, say for the favorite team, one could compute: favorite probability + 0.5*draw probability. The same could be done for the underdog team. In such a way, with no further assumptions we assume that the draw probability within the regular times is equally split between the two teams in the eventual extra-times. 

World Cup winning probabilities

We also provide some World Cup winning probabilities for the four teams, based on some forward simulations of the tournament.

The results are somehow surprising! Unlike for what happens for the majority of the bookies, Argentina has the highest chances to win the World Cup. France comes at the second place, whereas Morocco is the underdog, with only the 8% probability to become the World Cup winner.

Full code and details

You find the complete results, R code and analysis here. Some preliminary notes and model limitations can be found here. And use the footBayes package!

Final considerations

We had a lot of fun with these World Cup predictions, we guess this has been a good and challenging statistical application. To summarize, the average of the correct probabilities, i.e. the average of the model probabilities for the actually observed outcomes, is 0.41, whereas the pseudo R-squared is 0.36 (up to the quarter of finals matches).