06/13/2014 11:29 am ET Updated Dec 06, 2017

Here's What's Wrong With FiveThirtyEight's 2014 World Cup Predictions

This question originally appeared on Quora: What are some criticisms of FiveThirtyEight's 2014 World Cup predictions?
Answer by Yair Livne, Data Science Manager

These predictions don't look great:

  • Soccer tournament results are much more volatile than the predictions generated suggest -- so my first thought is that the standard errors are incredibly underestimated. If you look across different groups you'll see things like 90% probabilities for certain teams finishing at different position, which I find really hard to believe. For example, Brazil supposedly has a 94.8% shot of finishing first in their group -- it's enough that Mexico (a pretty solid team which can score a lot against weak teams) tie against them with a super defensive game to change this. That doesn't feel like a 5% scenario.
  • Historical performance in World Cups has been a great "feature" for predicting performance in future World Cups. For example, Italy has been able to achieve amazing success even with really weak teams over the years (due to awful, negative football), including recently when they made the finals in the Euro Cup in 2012. That seems to be totally absent from the current prediction which is based on some "quality" score. There's no way Italy has a 2% of appearing in the final, it's more like 10%.
  • Some things in the overall rankings just feel way off: England is placed far too high for a team that will never reach the finals, Chile's prediction is another example of relying too much on the very near history, the US won't get anywhere close to the final, Japan/South Korea seem massively under-ranked etc.
  • Sadly, and somewhat contrary to my first point, Brazil's chances of winning are perhaps underestimated. With the home court advantage (probably another missing feature from their model) and the level of corruption at FIFA today, I'd give them odds of >60% to win.
More questions on Quora: