The World Cup kicks off in a few hours, which means the results of the many, many prediction models that have popped up in the last few weeks will finally be tested. Below are some of the best examples; this is a good time to remind you that betting markets are usually the strongest predictors of all, if you're just looking to fill out a bracket. Here we go:
Bloomberg
Bloomberg Sports's World Cup page has a neat schedule and results tab in addition to the bracket, but unfortunately their model's methodology—which predicts a lot of 1-0 games—isn't laid out anywhere. Update: Bloomberg has sent along a link to a (partial) methodology; you can check it out here.
Who's going to win?
Brazil over Spain, with Germany and Argentina in the semi-finals.
What are the chances that the U.S. advance?
It doesn't look good. The USMNT are projected to tie Ghana and Portugal but lose to Germany, finishing third in Group G.
Does it have a cool visualization?
You bet! Same people who made the "How Americans Die" graphic that blew up the internet a couple weeks ago. Click on the individual games to see breakdowns of the matchup.
FiveThirtyEight
FiveThirtyEight's model is based on the Soccer Power Index, which Nate Silver created for ESPN back in 2009. SPI looks at national team performance, but it also, notably, tries to rate the players of individual teams based on their club-level performance. This isn't easy, but it does help alleviate some sample size issues; these national teams simply haven't played together that much. Detailed methodology here.
Who's going to win?
Brazil (45 percent chance), followed by Argentina (13 percent), Germany (11 percent), Spain (eight percent), and Chile (four percent).
What are the chances that the U.S. advance?
34 percent. Chances of winning? "<1%"
Does it have a cool visualization?
Yes. Instead of shoehorning the World Cup's unique format into a simple bracket, 538 made separate tabs for the "group stage"—a portion of which is shown above—and the "knockout stage."
The Economist
The Economist's team looked at the outcomes of every official FIFA game since 1993, and built a model based on the relative strengths of teams, when the game was played (more recent games have more weight), and where the game was played (home field matters). More detailed methodology here.
Who's going to win?
Brazil (21 percent chance), followed by Spain (7.6 percent), Germany (7.5 percent), Portugal (6.9 percent), and Argentina (6.4 percent). That leaves a pretty healthy 50.6 percent chance that one of these five teams won't win. This seems... generous, but who knows? Maybe they're on to something.
What are the chances that the U.S. advance?
Per this, the United States have a 22.7 percent chance (!) of coming in first in Group G, and a 27.2 percent chance of coming in second. The squad also has a 2.9 percent chance of winning it all.
Does it have a cool visualization?
It has a radial bracket that's tied to a bar graph. It's workable but a bit clunky.
Goldman Sachs
Goldman Sachs is an investment bank, and, apparently, investment banks like to show off their research departments by building World Cup models (see also: ING, Deutsche Bank). Anywho, fuck Goldman Sachs.
Who's going to win?
Brazil (48.5 percent chance), Argentina (14.1 percent), Germany (11.4 percent), Spain (9.8 percent), the Netherlands (5.6 percent). That just leaves 10.6 percent for everyone else.
What are the chances that the U.S. advances?
The U.S. have a 41 percent chance of making it to the second round, and a 0.5 percent chance to win it all.
Does it have a cool visualization?
Nah.
SBNation
SBNation's Michael Caley threw his hat into the ring for this World Cup, building out a clever model that's based on "the quality of chances that teams create or concede" instead of goals scored and allowed. A similar model predicted Liverpool's ascent in the Premier League this season; they had excellent chance-quality in 2012-13 but failed to capitalize.
Who's going to win?
Brazil (28 percent chance), Spain (24 percent!), Portugal (seven percent), Germany (five percent), or France (five percent).
What are the chances that the U.S. advance?
22 percent, with a 0.2 percent chance of winning it all.
Does it have a cool visualization?
It has some very well-formatted tables!
The Power Rank
Ed Feng's model, like all of his models, focuses on ranking teams with careful consideration to strength of schedule. His ranking algorithm isn't public (although he has a nice breakdown here), but the outputs have been fairly accurate predictors of soccer results, compared to some of his competitors.
Who's going to win?
Brazil (35.9 percent chance), followed by Argentina (10 percent), Spain (8.9 percent), Germany (7.4 percent), and The Netherlands (5.7 percent).
What are the chances that the U.S. advance?
15.9 percent to advance as the Group G winner, and 22.8 percent to advance in second place. 1.1 percent chance of winning it all.
Does it have a cool visualization?
Yep. Andrew Phillips of Chartball provided the visual here, another radial bracket with the bubble sizes proportional to each team's chances of winning. When you load the page, you'll notice that all these bubbles go flying into place. This is called a "force layout," and it's a cool-looking, usually pointless piece of design that every data visualizer secretly wants to use a bunch.
NumberFire
NumberFire is a generally excellent analytics site; beyond their World Cup "win odds," they also provide in-depth game predictions and betting tips. Plus, you can simulate games that probably won't happen in the World Cup; the U.S. would have an 11 percent chance of beating Brazil in a knockout game, for example.
Who's going to win?
Brazil (29.1 percent chance), Argentina (12.2 percent), Germany (10.9 percent), Spain (10.6 percent), or France (9.7 percent).
What are the chances that the U.S. advance?
25.8 percent, with a 0.15 percent chance of winning the Cup.
Does it have a cool visualization?
Not really their thing.