NWSL Playoff Probabilities and Predictions
/By: Arianna Cascone and Matthias Kullowatz
The 2022 National Women’s Soccer League Regular Season wrapped up earlier this month, and both the Shield and the playoff race came down to the wire. In the NWSL, the top six teams make the playoffs, the top two teams secure a bye to the semifinals, and the regular-season winner is granted the NWSL Shield along with a $10,000 per-player bonus.
Teams in third through sixth place battle it out in the first round, with the higher seed hosting. Semifinals are also played at the higher seed, and the NWSL Championship is played at a neutral site. This year, the last teams standing will duke it out at Audi Field in Washington, D.C.
With a nod to the ASA tradition of releasing playoff model predictions for Major League Soccer, we’ve worked out the odds that each team advances to the semi-finals and final, and also determine teams’ chances of winning the NWSL Championship.
Let’s talk first-round
In both matchups, with San Diego Wave versus Chicago Red Stars and Houston Dash versus Kansas City Current, our model favors the home sides. The baseline probability of a home team advancing between two evenly-matched teams was set to about 60% in the model, so there’s a little more to these predictions than the fact that the Wave and Dash will be playing in front of their home crowds. That is, the model’s key features are goal differential, expected goal differential, and receiving goals added differential (for more on the model creation, see Methodology below).
San Diego bests Chicago modestly in goal differential (i.e., +10 vs. +8 GD), but their receiving goals added differential is best in the league and a little over four times that of Chicago’s, and their expected goal differential is three times better than Chicago’s, too. The model favoring San Diego here could be a bit of a sanity-check, considering San Diego spent 11 weeks atop the standings and Chicago squeaked into the playoffs after the North Carolina Courage failed to win their final game.
In the case of Houston-Kansas City, the Current are ‘better’ than the Dash on all three of the key metrics included in the model, but the home-field advantage is still the biggest driver here. The gap between Houston and KC in the data produces a relatively small edge to KC before home field advantage is applied. Without home-field advantage, that means the math says Houston would have a 33% chance of winning in regulation, and a 48% chance of advancing. Since Houston’s the home-side here, their advancing odds are boosted further up to the 58% that’s listed above.
Onto the semis and finals
The OL Reign and Portland Thorns, two NWSL powerhouses, enter the chat when we move into the semi-final match-ups. These teams have 67% and 61% chances, respectively, of being in the NWSL Championship match this season, and the OL Reign have the edge over Portland in our model here for lifting the trophy. Realistically, though, I could see either of these teams being crowned NWSL Champions on October 29.
Portland boasted the best GD and expected GD of all teams this year, and if we’re thinking about the regular season specifically, their +25 GD was nearly double that of the OL Reign (+13 GD), who had the second-best GD. The Thorns’ impressive goal differential can be partially explained by Sophia Smith’s breakout season that saw her battle Alex Morgan for the 2022 Golden Boot and lead the league in goals added per 96’ (0.38). Granted, Bella Bixby’s nine shutouts this season, which was tied for first in the league with OL Reign’s Phallon Tullis-Joyce, definitely contributed.
Speaking of goalkeepers and shut-outs, the OL Reign only allowed 21 goals in 28 games, with 19 of those coming in the 22-game regular season. That’s 0.75 goals per game, which was best in the league. Portland was second-best here, with 28 goals allowed over 28 games. The top two teams are also in the top four in receiving goals added, remembering that San Diego topped the table on that metric.
While Chicago might not technically fall into the NWSL power-house category because they’ve never won a title, the Red Stars have made it to in their seventh consecutive postseason. They lost in the semi-finals from 2015 through 2018, and fell in the finals in 2019 and 2021. Obviously our model doesn’t know that, but since I do, if Chicago overcame the ASA odds and made it to the final this year, I wouldn’t be surprised. I also know that Mallory Pugh was third in the league in goals added per 96’ this season, so maybe I should tell the model that, too.
To wrap this up, I’ll just say that even though the model seems to really like OL Reign’s and Portland’s odds of making it to the finals, we shouldn’t sleep on the four other play-off bound teams. This is a league that recently witnessed last-place NJ/NY Gotham FC extinguish the first-place Portland Thorns’ hopes of winning the Shield on the last game of the season, despite the fact that said last-place team were on a league-record 12-game losing streak, shut-out in the five games leading up to their season finale, and needed three goals to tie up the match.
In other words, soccer in this league is chaotic enough that any team can win on any given day in the NWSL, and that won’t change in the postseason.
Methodology
While we’re working on replicating our MLS regular season and playoff simulations for NWSL, we’re using a simplified approach in the meantime, using the following steps:
Build a game prediction model on MLS data that predicts the probability of a home win in regulation. Key features in the model include goal differential, expected goal differential, and receiving goals added differential–three metrics we believe to be especially important to predicting future outcomes. The model is an elastic net logistic regression fit to the home teams’ binary win/not-win outcomes from MLS seasons 2013 - 2022, game numbers 22 through 34.
Apply the model to NWSL teams, using 28 games per team from both regular season and Challenge Cup Group Stage matches to get a bigger sample of performance.
Adjust the home win probability to more closely match historical NWSL home-field advantage. While MLS sees about a 53% home win rate, NWSL’s home win rate has been just 41% over the past six seasons. Assuming that home field advantages will be greater in the 2022 playoffs than in the 2016-2022 regular seasons, the baseline was set to 45%.
To adjust for extra time possibilities, assume that the draw-to-home-loss ratio is preserved from NWSL regular seasons, i.e., 28:31. Additionally assume that the probability the home team wins in extra time is between their probability of winning in regulation and 50%.
Example, a team with a 50% chance to win in regulation (22% loss and 28% draw) is given a 50/72 = 69.4% probability of winning, conditional on the game being decided in regulation. This is because regular season NWSL games are decided in regulation 72% of the time. Take the average of 50% and 69.4% = 59.7% as the probability the home team wins, conditional on going to regulation. Thus this particular home team has a 50% + (28/59)x(1-50%)x(59.7%) = 64.2% chance of advancing.
Neutralize home-field advantage for the final, played in Washington, D.C., as the Spirit are not participating.
Simulate the playoffs 10,000 times. While you could enumerate all 2^5 = 32 possible game outcomes, that is more annoying than writing a simulation.