2018-2019 Point Predictions
The NHL season is just about a month away from puck drop, rosters are taking shape, and expectations for teams have been set. Since I have put the time and effort into running various point projection models, I figure I might as well post them. Some of the projections I like, others I question. I’m not going to sit here and say that these are the best projections out there because the one thing I would bet on when it comes to these projections is other people will do a better job. However, since all point projections are guesses, some much more educated than others, I figure there is no harm in posting them. To be clear, don’t look at these projections and think you’re going to make a ton of money by beating Vegas odds.
The “Model”
These point projections were not pulled from thin air or by throwing darts blindly at a dart board (though that could turn out to be just as accurate), but through a simple regression model fed into a Monte Carlo simulation. Let’s start with the regression portion. We set our “y” as Regulation and Overtime Wins (ROW) and our x as Corsi For % during 5v5 play. The wins were then translated to points (multiplying by 2) and an additional somewhat arbitrary constant (16) was added to account for non-ROW points. To calculate each team’s Corsi For %, a weighted sum of player performance was used. The main question you should have as a reader is what the weighted sum of player performance is.
The answer somewhat complicated and a number lacking a little statistical rigor. We started by constructing projected rosters for each team and determining an estimated time on ice for each of those players. The estimated time on ice was then converted into a percentage of team total ice time. All of this was an educated guess, so flaws and bias definitely exist here. Once those were determined that percentage, it was multiplied by each player’s weighted Corsi For %. The weighted Corsi For % simply weighted last season’s performance by three, the season before by two, and the season before that by one. Players with less than three seasons of adequate experience had their performances imputed through an educated guess.
With each player performance determined and then summed to the team level, we then ran a Monte Carlo simulation, changing each player’s Corsi For %, using a normal distribution centered around their weighted average and standard deviation of performance. 90,000 instances were then run to determine the average point total for each team.
It is important to note that this “model” has not been properly back or forward tested so we can’t say how well this model should perform. This process is something on our radar to improve upon but creating a proper projection model this summer did not fall high enough on the priority list to develop one properly.
The Results
Below is a table showing the results for each conference. You can see who we have as playoff teams. Once again, we would strongly caution against taking our model to Vegas and expect to become rich. Once you look at the results, you can read on for my commentary on what teams we think will perform around our projections, and those that our model is too bullish or bearish on. Due to the nature of the model, the point spread is not going to be as great as expected. I would fully expect the top teams to have more points and the bottom teams to have less points.
A Little Commentary on…
Teams the Model is Bullish on
Let’s start with the teams projected at 99 points. I think Calgary is a team that will be better than they were last season but I don’t see them finishing at the top of the conference. The same with Carolina. Both of these teams were hurt by stretches of poor goaltending last season so seeing improvement is likely to be expected. Both teams have the talent to do it but probably not enough to be conference contenders.
The next team I think is much too high is Montreal. I think they would be a lot to go right for them to reach that level of point production. However, I don’t think it is impossible. A full season with Shea Weber and a bounce back from Carey Price could easily make Montreal a playoff team but if I were just making educated guess on points, would probably peg Montreal around 85 points.
The only other teams that stand out to me are Detroit and Vancouver. I think them, Ottawa, and both New York teams will likely be the ones battling for last place. Let me reiterate that these teams at the bottom of the standings are likely to finish with less points (maybe as many as 15 less), but I do think the order is relatively accurate. The model does factor in Zetterberg not playing this season. Vancouver is a team that will be hard to figure out as they don’t seem to think they are a bottom team. This model seems to reflect a point total that Vancouver’s front office thinks they can achieve, but I don’t see it. They are a team without great goaltending and without the Sedins, they are likely going to lack the goal-scoring necessary to win enough games to stay in contention.
Teams the Model is Bearish on
I’ll start with three of the best teams from last season, Nashville, Tampa Bay and Winnipeg. I was a little surprised to see both of these teams relatively low in the standings and fully expect them to be the class of the league again. These three teams just have too much talent to not be at the top. In likeliness of a drop from last season among those three teams, I would rank the teams as Winnipeg, Tampa Bay, and Nashville.
Minnesota and Colorado are both teams that will likely finish with slightly higher point totals, but I would not be surprised at all to see them finish outside the playoffs.
Teams that Readers Might Question
I don’t know how many of you actually will read this but if any Toronto fans read this and I don’t talk about them, I feel like I will hear about it. Yes, Toronto added John Tavares, yes, Auston Matthews will probably play more games this season, yes I do think Kyle Dubas has had a good offseason, but I don’t know if they are really going to finish with much more than the 95 points our model projects. I think there are still too many questions on the blueline and I am a fan of Gardiner, Reilly, and Dermott. However, I’m not sure if those three players will be able to play the tough minutes asked of them at the level needed for the Leafs to truly be one of the top teams.
The next two teams I need to address are two of the dominate teams of the early 2010s, the Los Angeles Kings and Chicago Blackhawks. The Kings I don’t think are as much of a stretch to finish that high with the addition of Kovalchuk and a strong returning roster from last season. The Blackhawks’ age is certainly catching up to them but they still have some of the most talented players in the league on their team. Our model does not factor in the potential of Corey Crawford not returning to form or not returning at all so that is likely the x factor.
Yes, this model has the defending Stanley Cup Champions missing the playoffs and yes, I have not discussed that yet. I would not be surprised to see Washington miss the playoffs. They are a team with a new coach who needed an elite stretch from their backup goalie last season to keep them in the playoffs. Couple that with a summer of a well-deserved celebration and the chance that might lead to a slow start and the recipe is there for them to miss the playoffs.
The final team to address is last year’s worst team, the Buffalo Sabres. Yes, the team traded away a top center in Ryan O’Reilly for depth pieces. However, the addition of the depth pieces should improve the team. Couple that with Jeff Skinner, Conor Sheary, and Casey Mittelstadt and I think they actually should be better overall up front, even without O’Reilly. Rasmus Dahlin should add the quality needed on the blueline and the team’s goaltending can’t be any worse. At some point the team has to take a step forward, so I’m ok saying this will be the year.
UPDATE: 9/10/18
I made a few changes to the model to better capture the extremes that are likely to exist. I also updated team rosters after the between Vegas and Montreal. Furthermore, I forgot that Shea Weber would miss half of the season so I adjusted the ice time of Montreal’s players accordingly. Most of the extreme smoothing was done purely arbitrary. The following teams had points subtracted: Ottawa, New York Rangers, New York Islanders, Vancouver, Montreal, and Detroit. The following teams had points added: Boston, Nashville, St. Louis, Pittsburgh, and Tampa Bay Lightning. These point additions were simply my logical choosing. Please note, only about half of the points subtracted from Montreal are the result of properly reflecting Shea Weber’s TOI and trade. These will likely be our final projections unless another major trade occurs (I’m looking at you Ottawa).
Updated 9/17 Post Erik Karlsson Trade
Hopefully this is the last time I have to update these projections but it is definitely necessary after all of the news out of Ottawa. They have lost Karlsson and Pageau and of course San Jose has gained Karlsson. Because of the nature of this model, the point impact of losing Karlsson and Pageau will likely be greater than this latest change reflects. The same can probably be said for San Jose.
If you would like to voice your opinion feel free to comment or tweet us (@afpanalytics).
All stats were courtesy of naturalstattrick.com.
KYLE STICH is the Director of AFP Analytics. In addition, Mr. Stich is a tax specialist and Director of Operations at AFP Consulting LLC, whose clientele include professional athletes performing services on three separate continents. Mr. Stich earned his Master of Science in Sport Management with a Concentration in Sport Analytics from Columbia University in 2017. He earned his undergraduate degrees in Accounting and Sport Management from St. John Fisher College in 2015, where he has served as an adjunct professor teaching Sport Finance and Baseball Analytics.