Let’s Experiment in EHM: Introducing a WAR-Optimized Roster for 2019-20

It was confirmed in this past week that the New Jersey Devils are done for 2019-20. They will not be one of the 24 teams returning to the rink as part of the league’s Return to Play plan. We will not see Devils hockey in action until the 2020-21 preseason, which may be in 2021 for all we know at this rate. This offseason, already over two months old, could continue for another six months. With a lot of downtime and not even any active hockey going on right now, we may as well find other ways to entertain ourselves.

I really appreciated the reaction to the Let’s Re-Do the 2019-20 Season for the Devils in EHM series that just recently wrapped up. So much so that I will continue to play Eastside Hockey Manager for this site and with more creative scenarios. One of the reasons why people play sports simulation games is to try things out that you would not be able to do in real life. Henceforth, let’s experiment with EHM.

Since the game has an easily editable database - thanks in large part to the utility program put up by archibalduk at the The Blue Line forums - we can try out a lot of things just to see how it will go. Maybe you want to see an all-European Devils squad or a team of 18-year old rookies or a team filled with players who weigh less than 170 points or whatever. Maybe you want to see the experiment take place in a junior or European league. Maybe you want an all-New Jersey lineup. I absolutely encourage suggestions for EHM experiments from you all, the People Who Matter, so please leave a comment if there is something you would like to see. And if it works out and it is interesting enough, then it will be a part of this series. Unlike the past series, I expect these to last for a post or two (unless more or needed), so if you do not like this experiment, then wait a week or two and there will be something else.

The Experiment

Replace the 2019-20 Devils roster with a NHL roster optimized for Wins Above Replacement based on the 2019-20 season. Play through the 2019-20 season with this WAR-optimized roster.

The Reasoning

One of the advances in hockey stats in recent years is Wins Above Replacement. As I understand it, the concept comes from baseball to identify how valuable a player is based on their contributions compared to a baseline, an easily replaced player. It has been a sought-after in hockey and the most recent and popular attempt came from the Younggren brothers. I personally do not think it is a good model. I think the model creates too many comparisons that really do not hold up in reality, such as claiming Valeri Nichushkin is more valuable than Art Ross Trophy winner Leon Draisaitl. What the model values is not necessarily so valuable and one of its primary variables is another model the Younggrens came up with, Regularized Adjusted Plus-Minus, which is pulled from basketball - which has its own limitations. I would suggest checking it out for yourself; while their explanation of their methodology is accessible, their results of the model costs money to access.

However, I am a tolerant man. I am more open-minded than the haters and losers, of which there are many, would like you to believe. I understand some have taken to this model like a duck to water. And while there are plenty of difficult conclusions from it - such as claiming P.K. Subban is a below-replacement player but Mirco Mueller is not - it does appear identify that plenty of the better players of the league are, well, among the better players in the league. A team of the best players by WAR would likely do very well in the NHL. To see if that is reasonable, we can see how a WAR-based lineup would do in a game that is meant to simulate the NHL. Eastside Hockey Manager is such a game.

In order to do this, I turned to someone who is a proponent of this stat. CJ, who is a part of this blog and therefore is a Person Who Matters, is very much pro-WAR. During April, I asked him to come up with a WAR-optimized lineup based on the 2019-20 season. He did an excellent job in solving what was effectively a linear optimization problem. I also asked him to provide an explanation of how he came up with the team. Here it is in own words:

Building the WAR-Optimized Lineup by CJ Turtoro

The question is “What is the perfect WAR team?” The answer to this question is going to be whatever combination of players 1) is cap-compliant, 2) fills the roster positionally, and 3) maximizes WAR or GAR. The mathematical answer to this question can be found using a technique called “linear programming” which is actually a concept from algebra. The basic idea behind linear programming is that you use all of the restrictions to create a “feasible region” — a list of legal rosters — then you find which of those teams has the highest GAR. If you’re interested in the math here is the link to my writeup. [JF Note: You should read it also because it is very good.] If not, I’ll describe the constraint matrices in layman’s terms here: 1) Combined cap hits can’t exceed $81.5M, 2) 23 or fewer players 3) At least 4 of each forward position, 6 defenders, and 2 goalies and 4) No player can be chosen more than once. Following those rules, there will be one “correct” team that maximizes GAR.

The team produced by this process had a cap hit of $80,402,021 — more than a million under the cap — and produced over 400 goals above replacement value. For some perspective on that figure, no real NHL team cleared 160 goals above replacement this season. It’s also not like this is a team full of Beau Bennetts and analytics darlings — these are universally renowned talents. In my opinion there are 6 top-line-quality forwards (Pastrnak, Marchand, Mackinnon, Panarin, Point, and Pettersson), 4 top-pairing-quality defencemen (Ellis, Josi, Slavin, and Makar), and a Vezina-quality goalie in Hellebuyck. Even with catastrophic injuries, I think this team should be historically dominant — less than 60 wins would be a letdown and 70 wins isn’t out of the question if we’re healthy.

The WAR-Optimized Roster in EHM

I took CJ’s team, used archibalduk’s database editor to make a separate database for this game based on xECK29x’s roster I used for Sherman Abrams’ and Trial Run’s games, and replaced the 2019-20 Devils with each of the players CJ identified. In order to prevent a run in the game on free agents, I swapped out the Devils with each player from their original team. (The game will not let me make 23 one-for-one deals.) For example, since Nathan MacKinnon is a Devil, I gave Taylor Hall to Colorado. This unfortunately meant I had to give some key Devils to Our Hated Rivals. After setting up the database file for the game, I created a new 2019-20 season with general manager Tom FitzGARald. CJ and I came up with the name. Here is the depth chart, per the game:

This is a very good team on paper. It is young as only Brad Marchand is the only man in his thirties. Darcy Kuemper and Roman Josi will hit the big 30 during the playoffs, assuming this team goes deep - which they absolutely can. As much Hischier and Hughes will be missed, the players on ELCs here are stunningly good from Cale Makar to Elias Pettersson to Anthony Cirelli. While EHM does not really incorporate WAR and some of the Devils removed ended up being among the better players in the game - and Hall being among the very best - there really are no scrubs here. I do not agree with CJ that the team will win 60-70 games, but I can see 50+ wins and a trip to the conference finals at a minimum. The board in the game thinks they are a Stanley Cup contender and that is something CJ and I would also agree with. This also means the expectations for success are high (the fans in-game are indifferent and likely are making fun of the capitalized GAR in the last name):

However, there are some limitations to make this experiment a little more difficult.

The Limitations

First and foremost, I have to use CJ’s roster as much as possible. This means I cannot make any trades with this team. This means I cannot sign any free agents. This means I cannot waive anyone. I have to use 20 out of these 23 players every night as much as possible. If it turns out that somebody stinks, or has a cold run of games, then I just have to deal with it internally. I can change tactics and training routines. I can scratch players as needed. I can set up the lines. But I cannot dump any of the 23 to Binghamton or offer them to any of the other teams or the waiver wire. I cannot sign a prospect or call one up to take a roster spot. This is my team as CJ put together.

Second, as this is a Devils blog, I wanted to use the Devils organization. This means I must use Binghamton’s system when necessary. The 23-man roster has two goalies, thirteen forwards, and eight defensemen. I can only call up a forward from the B-Devils if I have fewer than eleven forwards available. If I have eleven forwards healthy, then I must play them and go with seven defensemen provided I have seven defensemen. I can only call up a defenseman from the B-Devils if I have fewer than six defensemen available or unless I cannot play 11/7. Those who played EHM know that injuries are inevitable so there will be some games where this lineup of studs also has, say, Michael McLeod among them.

No disrespect to Binghamton, but there is a notable drop off in quality from this roster to the AHL roster. Especially in net. I did put Ty Smith there; but he is only to be used as necessary. That reminds me: I must send players down when the injured players are available to return.

Third, in order to ensure the first two limitations are met, I must play through the game as-normal. I cannot just “go on vacation” in the game and let it run. This is more personal and why some of these experiments will be two posts instead of one.

What’s Next & Your Take

The next post will be the results of this experiment. I’ll try to summarize what happens with the team in the season, the playoffs, and how the league in general went with the Devils on other teams as well as this WAR-optimized lineup.

In that post, I will also reveal what the next EHM Experiment will be. Again, please leave your suggestions for what experiment you would like to see tried out in EHM in the comments. Feel free to predict how well do you think this team will do in the EHM season. Will CJ be right as they win 60+ games? Will they go all the way to the Cup? What would stop them other than a crippling amount of injuries (entirely possible in this game)? Please leave your thoughts about the team as well as your ideas for future EHM Experiments in the comments. Thank you for reading.