clock menu more-arrow no yes

Filed under:

EHM Experiments: A Median in Name-and-Some-Stats-Only Squad

New, comments

This week’s Eastside Hockey Manager experiment is to put together a team that was in the league median in few categories. What would a team of 21 skaters with a expected goals for percentage between 50% and 51% and 2 goalies with league median save percentages of 91% look like? Find out in this post.

New York Islanders v Washington Capitals - Game Five
John Carlson: A xGF% in the 50th percentile, and much, much better elsewhere.
Photo by Chase Agnello-Dean/NHLI via Getty Images

Last week’s experiment was all about alliteration and how a squad of somebodies with the same initials for their name and surname could perform. It turned out well until Toronto exorcised some real-life demons in the world of Eastside Hockey Manager in the second round of the postseason. Still, it was a very competitive squad. The suggestion came from Ethmal1, who came up with another idea. What if you put together a team who was around the league average in a particular stat? How would they do? This is the subject for this week’s EHM Experiment.

The Experiment

Put together a 23-man roster of skaters who had an expected goals for percentage between 50% and 51% and goaltenders who finished in the league median for save percentage. See how they do in the 2019-20 NHL season.

The Approach

Ethmal1’s original suggestion was to use average points and save percentage. I opted to use the league median as that is more accurately representative of the middle. For points, the league median from 2019-20 is rather small since 883 players made at leats one appearance last season. It is 14 points. While the defense may be viable (and old, it would include Chara and Greene), it would be a really rough season. I would utilize the league median for save percentage to pick the goalies.

I swapped out production for expected goals for percentage as the expected goals model reflects how the run of play goes when that player is on the ice. Players who finished with a value of at least 50% but fewer than 51% would represent the “middle” of the league. They have had a positive impact but certainly not one that would stand out. More importantly, I would have more than 21 players to consider. Plus, we used Corsi and GAR in previous experiments, so I wanted xGF% to get its place in the Sun.

It turned out that the salary cap played a big role in driving this roster construction since some Big Name Players finished with a 50% xGF% but less than 51% xGF%.

Goaltenders (2): Alexandr Georgiev, Alex Stalock

Ethmal determined the league average was 90.9%. The league median was 91%. For the sake of the cap, I’m glad I chose the median as Price and Andersen are considerably more expensive than Georgiev and Stalock. For the sake of performance, well, uh, it is what it is.

Defensemen (7): John Carlson, Mattias Ekholm, Justin Faulk, Matt Dumba, Erik Brannstrom, Robert Bortuzzo, Brian Lashoff

Knowing that Washington was far from an analytics darling this season meant I had an ace in Carlson. Faulk and Dumba were rather expensive and, in retrospect, I probably would have opted for someone cheaper. Brannstrom was a good guess as a prospective player and helped out at the end by being cheap. For similar reasons, the depth was going to be a rotation of Bortuzzo and Lashoff. I would have to hope they would not be terrible.

Fun fact: Erik Karlsson was under consideration for this squad but the contract was literally too big to fit in. Especially with the options I had for wingers.

Forwards (14): Taylor Hall, Ryan Johansen, Vladimir Tarasenko, Tom Wilson, Evgeny Dadonov, Kevin Fiala, Justin-Gabriel Pageau, Alex Killorn, Dominik Simon, Dominik Kubalik, Emil Bemstrom, Oskar Sundqvist, Jared McCann, Joe Thornton

First, my options at right were limited to just Tarasenko, Wilson, and Dadonov. So they had to be on the roster. Second, it was tempting to taking Ovechkin or Kopitar, but I opted for cheaper options that I figured would be super-effective in EHM. Taylor Hall is as good as gold in the game and Johansen has always done fairly well. Third, I had to be more judicious with respect to the middle. Thankfully, Pageau’s extension would not kick in, and Fiala and Killorn are on great deals for what they could do. The depth was a mix of one known young guy to do well (Kubalik), one young guy who did quite well in real life (Bemstrom), a Blue (Sundqvist) and two Pens (McCann and Simon) that could play center in a pinch. Thornton was someone I wanted to pick in a past experiment and since these experiments often see the team do well, why not try to get Thornton a digital Stanley Cup?

Looking back after the experiment, I could have been smarter with my forward picks. I was not so up against the cap that I could have spent a little more, although I could not afford an Ovechkin or a Kopitar short of dumping Johansen and a cheap ELC player for one of them. But they were not the worst. As you will see soon enough.

The New Jersey Devils of 50 xGF% skaters and league median Sv% goalies. Also known as the Fifties.
The New Jersey Devils of 50 xGF% skaters and league median Sv% goalies. Also known as the Fifties.

I took the Devils roster, gave myself the same limits for a lot of these experiments (no trades, no waivers, only call up players if needed), named the GM after someone known for being Fifty (Curtis Jackson III), and started the season.

The Experiment Results: The Regular Season

The fans did not care much for Curtis Jackson III the GM spot, but the board expected a playoff spot. I figured they would get that. A first line of Hall, Johansen, and Tarasenko would make opposing defenders wish they had a pair of Depends on. Carlson can eat a ton of minutes. Provided the rest of the team does not collapse on itself, they could be one of the top 16 teams in the NHL.

After a nondescript preseason, the season began with Johansen getting a ‘4’ in a game. That was the result of getting thrown out on a game misconduct, which earned him a two-game suspension. But that was not the real issue with the squad. It was a rough first month with the team starting 5-4-1. Georgiev and Stalock were not all that good as both put up save percentages in the 88% range. There was plenty of internal movement to figure out who played well with who. It was not a terrible start, but it was not ideal.

The good news is that the Georgiev and Stalock started to put in better performances in November. The wins became more common and the Devils started to move up a bit in the standings. This hit a obstacle when Georgiev left the November 15 game with an injury. While the Devils beat Pittsburgh 5-4, Georgiev suffered a partially torn MCL and would be out for three months. My options in Binghamton were either Gilles Senn, who had a save percentage below 86% and recently had a rating of ‘3’ in a game, or Zane McIntyre. I called up McIntyre and hoped Stalock could take a heavy load.

It turned out that Stalock righting his own ship was not a mirage. He was not all that amazing, but he was playing better than he did in October. This was good as he received a majority of the starts - until I learned that McIntyre could do the job. At first, I kept him to only play in games where the Devils had a back-to-back set. After McIntyre did very well in those games, I gave him more appearances. Either to spell Stalock or to trust that the team could do well against a lesser opponent. I was right to do so and McIntyre ended up posting up a save percentage in the 91% range. Meanwhile, the offense was taking care of business with either Hall or Tarasenko on some kind of hot streak. The third and fourth line was more contentious with poor games happening more often. However, the talent at the top was enough to carry them. Just like the Ekholm-Carlson pairing carried whatever the third pairing was, although Lashoff and Bortuzzo were not heinous.

At the turn of the calendar, the Devils beat Boston on New Year’s Eve and that was the last straw for Bruce Cassidy. More relevant to this experiment, the Devils were first in the East with a record of 28-8-3 (59 points), led it by three, and had a goal differential of +49 (152 GF, 103 GA). The team was buzzing and the buzzing would continue through the 2020 portion of the schedule. The NHL All-Star Game came and off Hall, Tarasenko, and Carlson went. Tarasenko had a brilliant enough game to get the NHL All-Star Game MVP. And then they went back to work by playing Ottawa (NJ won 5-3).

Given the limitations of the experiment, February is not really a big month. Georgiev did return from his partially torn MCL in early February. He returned to the crease on February 11, only to leave that game part of the way through with another injury. This time it was a thigh strain, so Zane McIntyre and his surprisingly good performances only rejoined the team for two more weeks. Not that it mattered much as this Devils team kept on rolling and increasing their lead on the Conference well into March (Georgiev finally became healthy then). On March 12, the Devils clinched a playoff spot. They were 50-14-6 with a six point lead on Pittsburgh in the division. Hall and Tarasenko continued piling up the points but it would not be enough to seriously threaten for the Art Ross. No matter, the Devils finished the season with the league’s best record.

The final standings. The Fifties came in first.
The final standings. The Fifties came in first.

The roster turned out to go from strength to strength and never really had any real struggles. Sure, they ended their season at 5-5-0, but their spot was never in doubt. With 311 goals scored, the offense was fantastic. Yet, no one Devil was in the running for the Art Ross. Here are how the league leaders ended the season:

League leading scorers in 2019-20
League leading scorers in 2019-20

McDavid was just too hot for too long. Tarasenko came close but a couple of multi-week injuries kept him from being able to compile the points. Hall was healthier but he did not go fully off as he had done in past experiments. Fiala was often behind Hall in the lineup so as impressive as his 42 goals were - only McDavid and Gaudreau had more - he was not going to challenge for the league lead in points. Dadonov and Johansen were consistent scorers but it is tough to be a league leader when you shoot at 9.4% and 10.5%, respectively. Also: Dadonov took 360 shots in this season. That is stunning.

Although they did not make the top 24 in scoring, there was one leader from NJ: Carlson. Carlson led all defensemen in points with 15 goals and 52 assists for 67 points. This and his outstanding average rating of 8.52 (!!!) should make him a shoo-in for the Norris. Speaking of average ratings:

League leaders in average rating in the 2019-20 season
League leaders in average rating in the 2019-20 season

To put an 8.52 average rating in perspective, it was the best in the league. A case could be made that Carlson was the best player in the world in this simulation. Tarasenko was the closest to him and Hall and Dadonov both were members of the 8-club. Ekholm and Fiala just missed out. Yes, the Devils were a top-heavy team in terms of quality but that quality was golden. It was enough to carry the team.

The Devils by average rating
The Devils by average rating

While the Devils’ stars were carrying the team forward, there were some disappointments. Pageau and Killorn each had tendencies to have some poor performances on a somewhat consistent basis. Like one or two of them within a seven game time period. The depth players I relied on for the third defensive pairing and the fourth line were just not good. Bortuzzo and Lashoff were not the worst, but they were not that good and definitely not productive. Branstrom was better than both but not to a level where it was a massive improvement. Out of the forwards I chose for depth, only Simon and Kubalik ended up being consistently good. Bemstrom started off the season terribly but improved to be below average. Sundqvist had some big games surrounded by a lot of nothing performances. Ditto for McCann. And poor Joe Thornton. He was just not good. He played like, well, a 40-year old set to retire at the end of 2019-20. Which he was in the game; he announced it February.

As for the goaltending, the average ratings for Georgiev and Stalock were not too bad. However, their save percentages were a less-than-ideal 89.4% and 89.5%, respectively. McIntyre’s NHL stat line was 15 games, a 11-1-2 record, a 91.1% overall save percentage, and an average rating of 7.40. While I question if that would hold up for half of a season’s work, that suggests he could have been better than both.

Still, it did not hinder the Fifty-Devils from making the postseason. They stomped their way to a #1 seed. Could they go all the way? If only for Joe Thornton?

The Experiment Results: The Playoffs

The eighth seed ended up being a hot potato between Carolina, Buffalo, Our Hated Rivals, and Columbus. Carolina won that game to have right to take on New Jersey. This was a comfortable series for New Jersey. They won Game 1, 4-1 and Game 2, 4-0. While Carolina mustered up a 2-0 shutout win in Game 3, the Devils responded with a 5-2 win in Game 4. Game 5 was a decisive 4-1 win that saw the Devils through to the second round.

One of the teams to give this version of the Devils trouble was Pittsburgh. They won the season series 2-1-1 and out-scored them 16-11. Plus, they beat their hated rivals in the first round. The Pens could trip up these Devils in theory. The possibility increased when Pittsburgh won the first game, 4-2. But the Devils evened up the series in Game 2 with a 4-3 win. They took control in Pennsylvania with a strong 3-1 win in Game 3 and a 6-4 win in Game 4 featuring six goals within the last eleven minutes. However, Pittsburgh scored five unanswered goals in a 5-2 win in Game 5 to extend the series. The Devils did not risk any drama that would come from a Game 7, though. They pulled ahead and away in the second period in a 5-2 win to close out the series in six.

The opponent in the Eastern Conference Finals was Toronto. They had the best record in the Atlantic. They beat Tampa Bay in five games to get to this round. Mitch Marner, John Tavares, and Andreas Johnsson were going off on teams. While the Devils beat the Maple Leafs two games out of three in the season, this was not going to be an easy series.

Toronto straight up smacked New Jersey in the mouth in the first two games. They won 4-3 at the Rock to start the series off. In Game 2, they straight up embarrassed the Devils in a 8-2 game. Toronto took 24 shots and scored on a third of them. New Jersey was in a deep hole in the series, which was now moving to Ontario. Could they pull out a response? They would in Game 3 with a close 4-3 win. The series was tied in an epic-length Game 4 that required four overtimes. At 123:37, Dadonov continued his amazing playoff run with the game winning goal. The series was tied. And the Devils put Toronto on the brink in Game 5, making the fans at the Prudential Center feel much better after those first two games. They won that one 4-2. They had two chances to beat Toronto one more time and then play for Lord Stanley’s Cup.

Unfortunately, the Maple Leafs had other ideas. They scored the first five goals of Game 6, including four within the first ten minutes. New Jersey was shell shocked and could only put in consolation goals in a game that ended at 5-2. Game 7 was necessary. It was in New Jersey. And it was one for the ages. Killorn put the Devils up 3-2 just under a minute into the third period. But Kyle Clifford provided an equalizer with just under seven minutes in regulation. The score held steady and it would all come down to one shot deciding it all. Game 7 in overtime. And it did not last long. Marner finished a feed from Johnsson - the two playoff aces on Toronto - to end it all at 62:21. The Devils’ campaign was over. They just had to win one out of two games against Toronto and lost both.

Playoff tree - Toronto ended up winning it all.
Playoff tree - Toronto ended up winning it all.

Toronto would go on to win it all by dispatching a Winnipeg team in five games. The smugness out of Ontario and the hockey media festered for months. Their line of Marner, Johnsson, and Tavares just rampaged teams throughout the postseason. How did the Devils’ player stats finish up?

Devils players by average rating in the playoffs.
Devils players by average rating in the playoffs.

Just as with the regular season, Carlson, Fiala, Hall, Tarasenko, Dadonov, and Ekholm were really, really, really good. The postseason was a breakthrough for Dadonov, who just kept finding the scoresheet. His 23 points were only behind Tavares (30), Marner (28), and Johnsson (26) out of all playoff scorers. They showed up and did very well. While his average rating was undercut by that Toronto series (his save percentage was knocked down from the 92% range to 90.3% by the end of the Toronto series), Georgiev was up there after the first two rounds too. He took the crease away from Stalock, who was quite bad whenever he saw minutes in this playoff run.

The old adage for any playoff campaign in pro sports is that you need everyone to step up. Especially players who are not known for that. That did not really happen on this roster. Sure, Bortuzzo and Lashoff played better than they did in the season. But those are defensemen who played less than 14 minutes per game. The depth at forward was a real let down. Killorn only had four goals and was OK. Wilson just had the one goal and unfortunately took more penalties than shots on net. Kubalik was a ghost at times. Simon was very disapointing. Pageau had nine points but if he was not scoring, then he was not contributing elsewhere. It is one thing to see Thornton, Sundqvist, McCann, and Bemstrom not contribute much. But Simon, Pageau, Killorn, Kubalik, and Wilson were all better than this in the regular season. Had one or two of them performed better - especially in the Toronto series - then perhaps this team would have competed for a Cup. Alas, they did not. Joe Thornton would ride off into the sunset without one.

The Experiment Results: The NHL Awards

Did the team win any individual awards? Yes, they did. Here’s the breakdown of winners and finalists from this team:

  • Rocket Richard Trophy: Gaudreau (45 goals) (runners-up: McDavid - 44, Fiala - 42)
  • Ted Lindsay Award: Tarasenko (runners-up: Draisaitl, Crosby)
  • James Norris Trophy: Carlson (runners-up: Hedman, Reilly)
  • King Clancy Memorial Trophy: Carlson (runners-up: Tarasenko, Draisaitl)
  • Lady Byng Memorial Trophy: Lindholm (runners-up: Tarasenko, Gaudreau)
  • GM of the Year: Kyle Dubas (runners-up: Jackson III, Cheveldayoff)
  • Jack Adams Award: Sheldon Keefe (runners-up: Nasreddine, Maurice)
  • Hart Memorial Trophy: Ovechkin (runners-up: Tarasenko, Draisaitl)
  • NHL First All-Star Team: Carlson, Tarasenko (others: Andersen, Hedman, Ovechkin, Draisaitl)

I am pleased to see Carlson get the Norris as he was the league leader in average rating (King Clancy) and scoring by defensemen. I would have been shocked if he did not win it. I was also pleased to see Tarasenko get as much love as he did. I would have thought Carlson would have been a finalist for the Hart too, but Tarasenko was the Devils’ best forward by far and the in-game voters respected that. I am surprised to see that those were the only two to get any credit outside of Fiala, the coach, and the GM awards. I would have thought Hall could have made the second All-Star team. I did expect Kubalik to show up better among the rookies, but his season was not that good so it is what it is.

Experiment Conclusions

Admittedly, having an xGF% between 50% and 51% does not make one “average.” In real life and in the game, there were star players who fit that range, some more middling players who fit that category, and some players who barely played in the league. This made sense. The stat captures what the goal ratio would be based on the expected goal model, which is a result of all the activity that happens when that player is on the ice. It is not solely by that player anymore than CF% or SF% or some other on-ice metric. To that end, I do understand it may have not been the best way to identify a median player. While that made roster construction more interesting and a more competitive team, I wonder if I really captured the spirit of what Ethmal1 wanted.

The team did succeed in meeting the board’s expectations and exceeded them by taking first in the East. They went out to the eventual champions in seven-game series in the Conference Finals. It is not like they faceplanted when the postseason began. To that end, I am fine with calling this a success. However, I do recognize that the roster could have been better in some areas. I also recognize that a more holistic approach may have been needed to identify a league average or median player by these more advanced stats.

Thank You

Thank you again to Ethmal1 for providing another suggestion for this week’s experiment.

With the NHL Draft coming up in a few weeks and free agency following after, the EHM Experiments will be taking a hiatus. I do not know if or when it will return. It may be for a few weeks. It may be until after 2020-21 ends. We do not currently know when extra-long offseason for the real-life Devils will end. The NHL wants an 82-game season in 2020-21. We shall see whether the global pandemic has other ideas.

I want to thank everyone who has provided a suggestion for one of these EHM simulations, a reaction to the experiment in the comments, and/or just read this wondering what in the world EHM is. Eastside Hockey Manager is the game and it continues to be the stand-out hockey simulation game on the market. Thank you - the People Who Matter - for reading.