Many regular season and playoff probabilities are based on a combination of team offensive and defensive ability, how hard the schedule has been and will be, and other score effects like home ice advantage or injuries. But in the PWHL’s first ever season, we don’t have concrete numbers for any of those things.
So I tried out a different approach that I haven’t seen anyone take before to predict the end-of-season standings and playoff odds. Let me know if you have, I would love to see how other people have used this method.
First, we need a dataset containing team information, results, and end-of-season point totals. Therefore, we need complete seasons with consistent roster structures. Going back to 2010, I pulled 85 full seasons of pro women’s hockey, counting each CWHL, NWHL/PHF, and PWHPA team season as a data point. As a result of the above conditions, I unfortunately have to exclude the 2020-2021 PHF bubble season and all PWHPA seasons except for the 2022-2023 tour.
Next, I converted the team records to points using the current PWHL scoring system (three points for a win, two points for an OT win, one point for an OT loss, and zero points for a loss). I then calculated the points percentage at each game played for each team in their league. Points percentage is equal to the points earned by a team divided by the points they could have won. For instance, if my rec hockey team wins two games out of three in a tournament, they earn six points out of the available nine. Six divided by nine is .666, which is their points percentage.
Here’s where I hit a snag. Because of various factors (number of teams, funding availability, covid), many of the season lengths are different from each other, which makes apples-to-apples comparisons at points in a season difficult. To handle this, I calculated the percent of the season completed at each game for and then multiplied by 24, to artificially fit each season to the span of the PWHL season.
This next section talks through some math and models, so for those uninterested, feel free to enjoy the aesthetic of the graphs below and skip this paragraph and the next one. For those interested: now that I had a points percentage for each point of the season, I could construct a linear model to demonstrate the relationship between the points percentage at a given point in the season and the end of the season points percentage.
I used these models to get an estimate and standard deviation for each team’s final results based on where each team currently is in the season. Using these estimates and standard deviations, I sampled from the derived normalized distribution 10,000 times, or in simpler terms, I simulated each season 10,000 times. I used the results to construct probabilities of each team ending in each spot in the standings. This technique is also known as a Monte Carlo simulation.
I asked my favorite resident viz maker, An Nguyen, if she could turn these probabilities into a viz, and frankly, she knocked it out of the park. Check it out below:
Tell me if you’ve heard it before: Minnesota has dominated the PWHL’s inaugural season so far and are most likely to win the regular season. If you have, here’s your evidence: they have a 40.9% chance to finish first and a 93.4% chance to make the playoffs.
Don’t worry Montréal fans, the regular season championship is not out of your grasp yet. Montréal is actually favored to finish first when looking at their odds for each rank, but that gap closes more with each game.
Toronto’s in an interesting spot right now. They’re likely looking at a solid third place finish, but the next few games are pivotal to their chances of moving up in the standings.
Despite all of this, Boston and Ottawa are not out of the playoff conversation. Each team has a coin flip’s chance at fighting to keep themselves in the top 4 conversation.
And while New York isn’t favored to make it at this point, the odds are low, but never zero. For my own sake, I hope they can make it so I can catch some playoff games.
Hope this gives every fan something to dig into, and leave a comment or tweet at TIG if there are any comments or questions!
