Euros prediction (4)

I haven’t posted the results for the quarter-finals ahead of the games, but I did run the model and here were the predictions.

Not too bad, in the end: the model did get that Spain-Switzerland would be a tough one and the resulting 1-1 was one of the most likely outcomes. Similarly, while giving Belgium a bit of an edge, the model had predicted that their clash with Italy would be very close and the eventual 2-1 win for the Azzurri (yay!) was a rather likely outcome. Denmark were massively favourite to win (with over 60% chance over the simulations from the joint predictive distribution of goals scored), which was confirmed by the actual result, though the Czech Republic gave them a game, especially in the second half. The eventual 2-1 score was actually the second most likely, according to the model. Finally, while England were also tipped to win easily against Ukrain (50% chance of an England win according to the model), the actual outcome (4-0) was rather unlikely under the model — which I stand by: before the last game, England had only scored 4 goals in 4 games.

On to the semi-finals, I’ve updated the dataset with the results of the later games and re-run the model to predict the next two; the results are in the picture below.

Basically, Italy are slight favourite against Spain (though I think perhaps the match will be a bit tighter than that); this is, once again, very much in line with the bookies odds (who give Italy 20:13 — equivalent to a probability of winning of 40% — and Spain 11:5, or a probability of 31%). I’m kinda touching anything woody (I think Spain are not 2010 Spain, but they’re not 2021 Republic of Crappy Football either, so I sort of agree with the model, giving us a slight edge, but I know it’ll be a tough one!).

As for England-Denmark, my model slightly deviates from the bookies (who give England as clear favourites with odds against of 3:4, implying a probability of winning estimated at over 57%, to the 49:10 odds against Denmark, equalling a probability of only 17%). Leonardo’s model is closer to this prediction, while mine calls the game as very tight; England are slight favourite (41% change of winning) but Denmark are not far (39.5%) — this time I’ve used 50,000 simulations from the joint predictive distributions because I wanted to fully characterise uncertainty in the distributions and reduce Monte Carlo error.

I think England are probably favourite to win this game — but I all things considered, I don’t think the model is preposterous in giving the Danes a fair shout. Among the things that the model knows:

  • England will have the home advantage;
  • England are ranked higher in the FIFA league tables;
  • England have not conceded in the last few games;
  • Denmark are also ranked rather high in the FIFA tables and have done well in the past few games (apart from defeat against Finland).

The last point in particular, I think, is what has driven the consistent favourable predictions for the Danes that the model has spat out especially for the knockout stage (when they were hot favourites to beat both Wales and the Czechs).

Among the things that the model does not know:

  • England will have the home pressure — obviously the fans are over excited and this can be a double-edged sword (a bit like for Belgium against Italy in the quarter-finals… Some considered that game the “last chance for the Golden Generation” and I think some elements of “now or never” are already playing up for the English… This may or may not affect their performance, but then again, it may…);
  • Denmark have an extra bonding and togetherness coming from the horrible Eriksen incident. Obviously, they basically didn’t show up in the game that was played only an hour after that and Finland beat them, surprisingly. BUT: they then gave Belgium a massive run for their money (in the first half, Denmark kind of destroyed them — and the model prediction wasn’t suggesting a Belgian trashing of the Danes…). And they’ve gone from strength to strength since.

Given all this, while I still think England will probably be in the final, I think that the bookies are a bit overly-enthusiastic about it and, until tomorrow, I’ll stick with my model…

comments powered by Disqus