A Win Probability Guide To U.S. vs. Australia


Check out FiveThirtyEight’s Women’s World Cup predictions.

The U.S. women’s national team begins World Cup play Monday, opening their set of group-stage matches against Australia. Despite having drawn the “Group of Death,” the U.S. team is still heavily favored in each of its three Group D matches, and its odds of surviving to the knockout rounds look promising.

According to FiveThirtyEight’s forecast, the USWNT has a 68 percent chance of beating Australia (19 percent chance of a draw). But what if the U.S. goes down by a goal early? When should you start to panic?

Consider the following your in-game cheat sheet for the U.S. game against Australia — a guide to when to start worrying and when it’s safe to start looking ahead to the next match. Using play-by-play data compiled from recent seasons of top league play, we built an in-game win probability model for women’s soccer. The model calculates win/draw/loss probabilities as a function of game time, score and, most importantly, the pre-match odds.

When it comes to the availability of data in women’s sports, inequality reigns. When building a similar model for men’s soccer, I collected play-by-play results for 3,000+ matches while hardly breaking a sweat. The win probability model for women’s soccer is built from a much smaller sample, about 950 matches, and that data required significantly more effort to compile. As a result of this smaller sample size, the modeling shared here may be more prone to noise, and may be less precise than one built from a more robust data set.

Data issues aside, what can we do with such a model? For one, it can tell U.S. fans how safe an early lead is if a heavy underdog like Nigeria starts strong. Or, when the score is tied, the crucial point in a match when a draw becomes more likely than victory for the USWNT.

One of the nice things about being a heavy favorite is that victory is rarely out of reach, even if you’ve fallen behind early, or remain deadlocked in a draw late in the match. Let’s take Monday’s game against Australia. For a tie game, the USWNT’s chances of defeat hold relatively steady for most of the match. With loss probability largely fixed, the U.S. is effectively trading win probability for draw probability the longer the match remains tied. At about 62 minutes in, a U.S. win and a draw are equally likely, and after that a draw increasingly becomes the most likely outcome.

beuoy-datalab-australia-1

What if the U.S. falls behind? When trailing by a goal, the match probabilities develop in a much different fashion. The draw and loss probabilities practically move in lockstep until about 75 minutes in, when they shoot off in opposite directions. It seems that the 75th minute marks a crucial turning point for many matches, at least according to our 950-match dataset.

beuoy-datalab-australia-2


Bonus Hot Takedown Podcast: FiveThirtyEight’s Allison McCann talks with ESPN’s Julie Foudy and USWNT players Kelley O’Hara and Christen Press. Subscribe on iTunes.