With the end of the group stage of Euro 2020 it is time to check on how the model is doing, and look at the updated new probabilities for who will win the tournament.
The description of how we validate the model can be found here.
As well as the models described in that post I have also added a new model “Average” which is, unsurprisingly, the average of each of the other model predictions (excluding ‘Lazy’ – in these kind of model averages you shouldn’t include something which is consistently the worst model).
| Model | Brier Score |
| HAL 9000 | 0.518 |
| Average | 0.521 |
| Bookies | 0.540 |
| SSC | 0.567 |
| BDC | 0.608 |
| Lazy | 0.667 |
On conclusion of the group stage our model is still winning, followed closely by ‘Average’ and then the Bookies. So I am pretty happy with that!
So, using our model (that we definitely have confidence in now!) we can look at updated probabilities for teams winning the tournament:
France are still the favourites for the tournament, but Italy are now 2nd favourites (which given their recent form, is no real surprise!). The largest (positive) change is for Netherlands, which I imagine comes from being drawn on a favourable “side” for the knockout rounds.
England are still only 6th favourite, but the model does give England a (slight) advantage for the match against Germany (41% chance of victory).