Banks are good at making money (or crashing the economy), but can they predict the outcome of the World Cup?
The short answer: kind of.
On May 30, Goldman Sachs released its 67-page World Cup and Economics 2014 report, that—between interviews with Franz Beckenbauer and in-depth breakdowns of every team (really, it’s a pretty good read)—includes a statistical model designed to predict the outcome of the tournament.
The model was developed by the folks at Goldman’s Global Investment Research division who, with a “purely statistical” approach, analyzed 54 years of mandatory international matches— that’s over “14,000 observations”—to determine the winner of the World Cup. The model uses a regression analysis, or a way to forecast the relationship between a number of variables, including a country’s Elo ranking, the average number of goals scored by a country over its last ten international matches, whether a country has a tendency to systematically outperform—or underperform—at a World Cup, and homefield advantage, among others.
Put all those criteria in a blender...and what’s that final match look like? Brazil 3-1 Argentina, according to the first edition of the report. Here’s the the full prediction for the knockout rounds:
What statistics couldn’t predict, though, was that this World Cup would be an absolute blinder: more goals have already been scored in Brazil than the entire 2010 tournament. The group stages brought us herculean individual performances and unimaginable upsets, and the knockout rounds promise to deliver more of the same. If you look at the chart above, you’ll see some discrepancies from the tournament's actual standings going into tomorrow’s quarter final clashes—notably the advanced positions of Spain, Portugal, Russia, Ecuador, Italy, and Iran.
Goldman, of course, has adapted. Following the conclusion of the Group Stage, analysts reran the formula, using updated statistics to re-predict the outcome of the tournament. Here’s the latest:
And here’s another chart, which highlights the picks that Goldman did get right.
The most interesting change in this updated model is the World Cup final—why, despite Argentina’s stellar performance throughout the group stages (and first knockout round) is the team downgraded? Why is the Netherlands now primed to lose in the final?
“The reason that the Netherlands are projected to reach the final instead of Argentina is their “Elo rating” rose as a result of their strong group stage performance (mainly by beating Spain),” wrote Jari Stehn, one of the authors of the report, when asked for comment.
Because the Elo rating is based on a country's entire historical track record, the Dutch—already one of the model’s highest rated nations—gained a slight advantage over their Argentinian competitors with their rampant group stage showing.
Stehn went on to explain that while the model is rerun after every new set of results, the ratings aren’t designed to change very quickly. “It make makes sense to allow for some impact from what you call ‘freak’ results because it just may be that they contain important information about a team’s true quality and it would be a mistake to ignore that,” he wrote.
Will the predictions hold up? If the first half of this World Cup is any indication, it’s anyone’s guess.