Applying actuarial and data science to (so far) politics, and sports
Thanks for reply. Another issue with the model is the idea of discovery over time. It could be that Costa Rica is quite good and we dont know it yet because their players are just now developing or they are reacting to the weather in unexpected ways, or they have a new strategy that is better suited to the world cup. The conditional probability of them winning a semi-final shown here is around 20%, but if Costa Rica manages to get to the semi-finals, certainly the current Vegas opinion of their strength will improve. Whereas the Vegas opinion of Netherlands is unlikely to improve as much if they make it to the semi-finals. So this model would tend to underrate low-rated teams' odds of deep advancement. In my model I add an error term that randomizes each team's 'undiscovered' strength, and this error term decreases over time as teams play and reveal their true level of world cup skill. Another good example of this is Uruguay. Oddsmakers have them as heavy underdogs to England due to Suarez' fitness uncertainty. But his fitness in the England game is highly correlated to his fitness in the semi-final round. They will likely only get to the semis if Suarez is fit. This is why Vegas has Uruguay at 60-1 to win the tourney and Ecuador with similar probability in your model, is 250-1.This is a cool website and fun model, this is not meant to disparage your work. Especially the 3x3 conditional outcome graphics, I really enjoyed it. Oddly enough I'm also an actuary named Patrick wasting huge amounts of time modeling the world cup.