How to predict the World Cup Winner?
With the World Cup just around the corner we thought it would be a good time to take a look at the ways that cloud technology, data and devops have been entering modern sport.
Using statistical analysis to attempt to forecast sports outcomes and performance is hardly new in 2018. American statistician Nate Silver, who rose to prominence with his accurate forecasting of the 2008 & 2012 US Presidential elections, took a stab at predicting the 2010 World Cup. Despite the sophisticated mathematical modelling, it was outperformed by Paul the Octopus.
Still, it’s safe to say that the methodologies have come a long way since then, with the use of machine learning and access to a wealth of data playing a key role. Similarly, there are far more people and teams attempting these forecasts, either for personal or promotional benefit. Data company Gracenote are backing Brazil with a 20% chance to win the tournament. Merse-fronted analytics firm Alteryx’s model gives the edge to Germany, as does data science blog KDNuggets.
EA Sports
On the other hand, EA Sports boldly predict that France will beat Germany in the final after a penalty shootout. EA’s internal search for the developer who forgot to program “Germany doesn’t lose on penalties” into FIFA’s World Cup mode is ongoing. Slightly related - we found a great article showing how EA has been using DevOps and automation to improve their development process , (particularly cool to hear they use Jenkins too).
Panini Stickers
The Toulouse School of Economics has taken a novel approach, concluding from 4000 Panini sticker albums that “Teams whose players looked angrier or happier performed better in the group stage of the World Cup compared to more inexpressive teams.” So if you have the average £800 required to fill your album you may be able to get a competitive edge over the bookies!
Big Banks Getting Involved
Even the big banks have got the tape measures out for this one, with several releasing World Cup forecasts in the last week. Japanese bank Nomura picked France, UBS goes for Germany and ING have their money on Spain. But Goldman Sachs definitely have the “yugest” model, simulating 1 million variations of the tournament to conclude that Brazil will beat Germany in the final.
A Goldman analyst is shown here searching for the simulation where England win the World Cup:
If you’re interested in learning more about the internal workings of these models, Gerald Muriuki has a great post that explains the building of a machine learning model to predict the outcomes of every match. He also ultimately hands the greatest probability of victory to Brazil.
Cloud Gatorade
And if Brazil do return from the tournament victorious, it could be down to two words: Cloud Gatorade. Gatorade has been rolling out Gatorade Gx, a cloud-connected water bottle system that will allow real-time monitoring of the players’ hydration levels. No doubt the Gatorade security team will be on high alert if Brazil wind up facing Russia in the tournament.
Data is the winner
Germany’s 2014 World Cup win has been credited to its effective use of data, and this time around it’s likely many of the teams will be using some form of cloud-based real time analysis to try and gain a competitive insight. This case study from Intel demonstrates just how far the tech has come, now being utilised by Bournemouth and Brazil alike.
Statistical models vs Pundits
Nonetheless, I can’t help but feel that all of these will fall short of capturing the unpredictable magic of an international football tournament that I hope we have to look forward to.
In fact the most pressing question I’m left with after looking at all this analysis is: would any of these statistical models be able to out-predict Paul Merson?