The final two weeks of the PWHL season will be unlike any in the history of major professional sports.
The New York franchise is on the brink of being eliminated from the playoffs, and could officially be out of contention as soon as Wednesday night. After that, the Gold Plan goes into effect.
Long discussed as a radical solution for tanking, the PWHL officially announced it would adopt the Gold Plan for determining its draft order in February.
GO DEEPER
PWHL sets another women’s hockey attendance records
The concept goes like this: Instead of a lottery system that encourages losing by awarding the best odds to the league’s worst teams, a league would determine the draft order based on the number of points each team earns after being eliminated from the playoffs. The idea was first introduced at the MIT Sloan Sports Analytics Conference in 2012 by Adam Gold, a PhD candidate at the University of Missouri.
It gained a following among fans and media, including The Athletic’s Sean McIndoe, who has been beating the Gold Plan drum for years. In 2016, Shane Doan suggested the same idea, after seeing his son, Josh, cheering against the Arizona Coyotes in the hopes of the franchise winning the lottery and the right to draft Connor McDavid the previous season.
“The second you start convincing someone to cheer against you, you’re one step closer to losing them as a fan,” Doan said at the time.
In the PWHL, once a team is mathematically eliminated from the playoffs, it begins earning “draft order points” in all subsequent games using the league’s standard points system. The team with the most “draft order points” at the end of the regular season will be awarded the first overall pick in the draft. The second non-playoff team will have the second overall pick. Playoff teams will select three through six based on the inverse order of the regular-season standings.
So, the more games New York wins after being eliminated, the better their chance at the top pick — likely Team Canada and Princeton star Sarah Fillier — becomes. Simultaneously, the league’s top teams (Toronto, Montreal and Minnesota) will be jostling for the No. 1 seed in the playoffs to be awarded the opportunity to select its first-round opponent between the third- or fourth-place teams.
Before the Gold Plan officially begins, The Athletic caught up with Gold — now a 40-year-old data scientist for Bayer Crop Science in St. Louis — about his model, the PWHL and what’s at stake over the final 10 games of the season.
Some questions and answers have been lightly edited for length and clarity.
Can we start with how you got the idea for a mathematical elimination model in the first place?
It was conceptualized back in March 2008. I’m a big St. Louis Blues fan and this was the year of the Steven Stamkos draft. That year, a lot of Blues fans wanted the team to try to lose to get Steven Stamkos and would say, ‘Let’s tank, let’s trade everyone, let’s just do everything we can to get Steven Stamkos.’ And I remember (former Blues President of Hockey Operations) John Davidson would be on Fox Sports Midwest telling Blues fans, ‘You don’t have to cheer against us for a draft pick.’ I just thought he shouldn’t have to do that and there had to be a way to figure out the draft order.
I thought, why can’t I do it? I have a math degree and then it hit me: Mathematical elimination. I lost sense of the world in a way. I was just like, Oh my God, I have to start to write about this.
You first presented this as a solution to tanking at the MIT Sloan Sports Analytics Conference back in 2012. What was the response like back then?
(Laughs) Well, you either love it or you hate it. I feel that the people who hate it, don’t fully understand how much flexibility there is. You can award only the first pick to the team that performs best after elimination, so that way the team with the worst record gets no worse than pick No. 2. But the plan would still encourage them to try until the end of the season. The biggest thing is you can calibrate the requirements for the league.
Why is mathematical elimination better than a draft lottery?
There are two different variables to optimize: You want to maintain competitive balance and give the teams that need the most help the best chance to get the best pick to get the best player. Simultaneously, you want fans to be able to support their favorite team, so that even the teams in the bottom of the league are assets. If you showed a team’s full schedule of games to a fan, they would want to say, ‘There’s not a single game we want to lose.’ And this draft order meets this requirement.
For the PWHL, there were only 24 games and six teams with four making it to the postseason. If you had a lottery, it might be a 50 percent coin flip for the first overall pick. But if you write out a simulator for mathematical elimination, and you actually make predictions on all of the games, you can beat that coin flip and award the first overall pick to the team that finishes worst, more frequently than a draft lottery could.
Can you take me through how the PWHL ended up adopting this plan?
Back in January, Burton Lee (a member of the PWHL’s hockey operations department) reached out on Instagram actually. And I couldn’t sit still for weeks. I was like, Oh my God, is this real? Eventually, I got a chance to sit down with Burton, and we talked about whether the PWHL was going to adopt this, and eventually they announced it on February 28.
GO DEEPER
PWHL announces playoff format, Gold Plan draft-order system
In the time between January and February, I built up a simulator for the PWHL that made estimates on all kinds of things. What’s the probability of awarding the first pick to the team with the worst record? How many games will teams play after elimination? And we are on par with everything that was predicted. Right now, the model estimates that the team with the worst record (New York) has an 88 percent chance of picking first overall. If we go back to when the season started, it was around 62 percent.
The expected number of games played after elimination the simulation predicted was 3.3. If New York becomes eliminated on Wednesday — I’m predicting a 96 percent chance of that occurring — it will be in line with the simulations and the output that it provided for the PWHL.
And I just have to say, I went to the MIT Sloan Conference and I had 15 minutes and I presented the best case I could. And it’s a credit to your colleague Sean McIndoe and other writers for keeping this idea within the public domain and keeping it active in the minds of sports fans. I don’t know if the PWHL would have been adopting this if it wasn’t for Sean and his persistence in writing about it.
I’m glad you mentioned New York. They’ve struggled to string points together this season, Is it possible for a team like Boston, currently in fifth, to ultimately win the first pick?
Boston and Ottawa are still contending for that final playoff spot. The difference between pick No. 1 and pick No. 3 is not worth losing the opportunity to become champion of the PWHL and lift the Walter Cup. New York has unfortunately fallen back a little bit and is almost always going to pick first in my simulations, with an 88 percent chance.
You could pick a team (on the interface Gold built for the PWHL) like Boston and the estimates will tell you the probability that they finish fifth and pick first. So they could take that first pick from New York after trying to challenge Ottawa and failing, but it’s currently less than a 12 percent chance.
If New York is eliminated on Wednesday, that sets up an interesting game on Sunday against Toronto …
The probability that Toronto can clinch the No. 1 seed in playoffs and pick their opponent goes from 54 percent if New York wins in regulation up to almost 90 percent if Toronto wins in regulation. So Toronto really wants to win this game in regulation because it’s almost going to guarantee that they can pick their opponent from whoever finishes third and fourth. But so does New York so they can win the first-overall pick. This is just such an extreme dynamic that we don’t see in any other sports league.
It definitely adds juice to an otherwise meaningless game between the best- and worst-ranked teams in the league. That’s the goal with this plan, though right?
We will be seeing an opportunity for Alex Carpenter and the PWHL New York team to be able to earn Sarah Fillier. Can you imagine if you’re at that game against Toronto and Sarah Nurse has a chance to stop New York from getting the first pick? Can you imagine the rivalry it could start if New York wins the chance to get Sarah Fillier and Toronto does not get that high probability of clinching first overall? It’s going to be incredible to watch.
The PWHL is going to make it a demonstration for all sports leagues that this model can work. They’re trailblazers, and they’ll be able to claim that they were the ones who ended tanking within sports leagues.
On the other hand, what’s the most common objection you hear about this model?
What happens if a team can’t win? I think it’s a really interesting discussion point. Looking back at the NHL, there are really only two teams that have challenged that over the last 10 years. The 2019-20 Red Wings won 17 games and the 2013-14 Sabres won 21 games. Even this year’s San Jose Sharks (19 wins) are an example.
Let’s say that it was once out of every five seasons that a team cannot win games: You shouldn’t have a draft order be primarily determined to identify that team and give them the first pick. They should still be awarded a top pick, which can happen using a flavor of my model. But you need to be able to maintain a competitive balance by keeping fans interested and engaged in their favorite team every game of the season.
In an alternate universe where the NHL’s draft order was set by mathematical elimination, how would it play out this year?
The first overall pick would have gone to the Chicago Blackhawks. The Blackhawks, the Sharks and the Ducks had really poor seasons. When Connor Bedard came back, it reinvigorated the team and they started winning. And when they did, they played themselves out of having the highest probability of winning the first pick with the lottery.
With my model, I think that the story down the stretch could have been: Can any team stop Chicago? If you stop Bedard now he does not get Macklin Celebrini and it would have been a very different narrative. It would have presented a big opportunity for Arizona, too. If they had three more points they could have gotten that first pick instead.
I actually pulled the data for Arizona and I want to mention a few picks they could have had with mathematical elimination over the last 15 years: Victor Hedman (2009), Nolan Patrick (2017), which was the pick that was used by Philadelphia over Miro Heiskanen, Cale Makar and Elias Pettersson. And, in 2018, Andrei Svechnikov. Arizona would have had four second-overall picks, and the highest they picked was fifth overall in that span.
(Top photo of the PWHL New York bench during a game: Bruce Bennett / Getty Images)