Google Cloud Platform is 11 for 12 in World Cup predictions

But the explanation for the wrong Germany prediction seems a bit weak in my opinion. Shots/Shots on target should not be a single deciding feature - there are many other featurs which have an influence on "expected goals" and which may have been in Germany´s favor.

The high certainity before (69% France) seems a bit off, especially compared to the betting markets (c. 69% Germany). Additionally I wonder about the high probability for a win vs brasil now (even without incorporating the loss of Neymar and Thiago Silva) - did they do a recalibration or change the model?

I would love a bit more background, like in 538s excellent piece about Messi[0].

[0] http://fivethirtyeight.com/features/lionel-messi-is-impossib...