Has Barry Trotz Hurt The Islanders This Season?
Many NHL media members, fans and especially broadcasters consider Barry Trotz, head coach of the Islanders, the best bench boss in the business.
Especially broadcasters? Yes, oddly enough, the Jack Adams Award for best coach in hockey (which Trotz won in 2016 and 2019) is voted on by the NHL Broadcasters Association.
Being recognized as the best in your field is great (speaking from personal experience, of course), but naturally, it comes with pressure and expectations. Once you’ve been recognized as the king (or queen), that’s the new bar. You’re no longer given the benefit of the doubt — there are no more chances to prove yourself.
Today we’re giving Trotz what he’s earned – a critique of his performance this year. Analyzing his in-game coaching decisions would be cool, but unfortunately, we have neither the time nor the context. We will content ourselves with analyzing his lineup decisions this season – what’s worked, what hasn’t and what might if he just gave it a chance.
Without further ado, it’s time for the second edition of Analytics Toolbox!
Individual, On-Ice, and Expected Goals Percentage
In the first edition of analytics toolbox, we focused on Kyle Palmieri and his expected goals. Today, we will continue our focus on expected goals – remember, it’s my favorite metric – but extend the discussion past the evaluation of only a single player’s scoring chances.
If expected goals models’ single utility were to tell us if a player has had rotten shooting luck, they wouldn’t be all that useful. We already have good old-fashioned shooting percentage for that anyways.
Fortunately – with a little help from our good friend’s addition and division – expected goals models can tell us a whole lot more! To demonstrate what else expected goals models have to offer, we’re going to look at Palmieri again. Josh Bailey and Jean-Gabriel Pageau, his two most common linemates this season, are coming along for the ride as well.
The Palmieri-Bailey-Pageau combination has 5.3 expected goals for and 7.2 expected goals against in 152 minutes together this season. That shakes out to a 42.4% expected goals percentage. Let’s break down what those stats are is telling us, step-by-step.
First, the 5.3 expected goals for. Last week in our discussion on Palmieri, we referenced his individual expected goals. Those are expected goals generated from his personal shot attempts – the puck coming off his stick. The 5.3 expected goals for that the Palmieri-Bailey-Pageau line has represents their on-ice expected goals. This is the sum of all Islanders’ players individual expected goals when they are on the ice. That’s the sum of the three linemates’ (Palmieri, Bailey and Pageau) individual expected goals plus the individual expected goals of the defensemen playing with them.
Obviously, the Islanders aren’t the only team generating scoring chances. The 7.2 expected goals against for the Palmieri-Bailey-Pageau line represent their opponents on-ice expected goals. You may have noticed that 7.2 is larger than 5.3. Yup. We’ll discuss if that’s good (it’s not) and how we use these number for analysis in a little bit. Before that, I do have to explain the last metric I mentioned above: the 42.4% expected goals percentage.
Expected goals percentage is calculated as on-ice expected goals for divided by the sum on-ice expected goals for and on-ice expected goals against. It is quite literally the percentage of total expected goals a players’ (or lines’) team generates when they are on the ice. The Palmieri-Bailey-Pageau line’s 5.3/(7.2+5.3) = 42.4% expected goals percentage reveals that when they are on the ice together, they are giving up more expected goals then they are generating.
Another advantage of the percentage form is that it allows analysts (or fans like us) to quickly compare performance compared to other lines and players. For example, the Martin-Cizikas-Clutterbuck line has played 378 minutes together this season and owns a 50.5% expected goals percentage. That is eight percentage points higher than Palmieri, Bailey and Pageau.
Okay. So we know how all these numbers are calculated, great. What’s the point? Why do we even care about on-ice expected goals? I’m glad you asked. Remember the definition of expected goals I gave in the first analytics toolbox?
“Expected goals is a model-based metric used to isolate the evaluation of play-driving and chance-creation/suppression ability from things a player cannot control such as bounces, quality of goaltender, etc.”
That quote sum’s it up nicely. On-ice expected goals are useful for assessing how effective a player (or line) is at generating – and suppressing the oppositions’ – offense, irrespective of if that offense ends up resulting in actual goals. An on-ice expected goals percentage above 50% is generally considered “good” because it indicates you are creating more chances than you give up.
The basic hypothesis: if a team consistently generates more expected goals than they give up, they will win more games than they lose.
The data supports this hypothesis: check out this Instagram post of mine from January.
The figure in the post shows regular season points percentage (the percentage of points a team earns out of the 164 available points they would earn if they won all 82 games) vs. expected goals for percentage. The relationship is abundantly clear — teams with larger expected goals for percentages are more successful.
Alright. I hope I’ve done a sufficient job explaining how these metrics work and convinced you that they are meaningful. Now, we are going to put them into practice! It’s time to talk about Barry Trotz and some of his coaching decisions this year.
Bringing it Back to Barry
Let’s start by looking at the most common Islanders’ lines this season. All forward line combinations with more than 150 minutes together at 5v5 are shown in the following table from MoneyPuck.com, sorted by expected goals percentage.
This immediately allows us to see what has worked and what has not. The most dominant line by a wide margin is the combination of Parise-Pageau-Oliver Wahlstrom, with an increase in almost 8 percentage points vs the next best line. In fact, only two other lines have expected goals percentages above 50%. That’s not great.
The percentages are calculated from the data in the other other columns of this table. A couple interesting things here – while the Parise-Pageau-Wahlstrom line is the most dominant, they don’t generate the most offense – the Anthony Beauvillier-Brock Nelson-Anders Lee combination has a slight edge in that regard (look at the xGoals For Per 60 Minutes column). Unfortunately, they are much weaker defensively, giving up almost an entire additional expected goal against compared to Parise-Pageau-Wahlstrom (xGoals Against Per 60 Minutes).
I want to take a moment to discuss a topic that has been a sore point for a lot of the Islanders’ fan base this season: Mathew Barzal and Wahlstrom playing on the same line. This finally started happening with some regularity in March, though many think this is too little, too late. Based on expected goals percentage you could argue Barry Trotz is justified in keeping them apart for so long. After all, the Islanders’ most dominant two lines this year feature one of each of these players.
I believe that interpretation of the situation misses the mark, and the xGoals For Per 60 Minutes data can help us see why.
The Parise-Pageau-Wahlstrom and Lee-Barzal-Bailey lines rank 65th and 67th respectively in xGoals For Per 60 Minutes among forward lines NHL-wide with at least 150 minutes of ice-time. Not good. Even the Islanders’ best line in this metric, the Beauvillier-Nelson-Lee combination, ranks 64th league-wide. To put this into context, this means every other team in the NHL has at least two lines (on average) that generate offense more effectively than any line on the Islanders. That is a problem. Playing the majority of the season with line combinations that have difficulty getting anything going offensively compared to your opponents is not a recipe for success.
Why didn’t Trotz try the Barzal and Wahlstrom combo sooner? Saddling a player that Elite Prospects calls “a sensational scoring threat” to Jean-Gabriel Pageau doesn’t make any sense and probably hurts his development. The fact that the Islanders’ have Barzal, one of the most creative playmakers in the league, makes this especially true. Sensational scoring threat + elite playmaker = a no-brainer to me.
The numbers back this up. In 134 minutes together at 5-on-5 the Parise-Barzal-Wahlstrom line is generating 3.67 xGoals For Per 60 Minutes. That is greater than one goal per 60 more than the previous combinations! Granted, they are giving up a bit more, but they still own a 53.2% expected goals percentage. The best way to build their defensive chemistry is to continue to play them together. Players of Barzal and Wahlstrom’s talent will figure it out.
One final statistic to note: their actual goals for percentage is 75%. That’s fantastic. Regression can certainly be expected, but not to the 53.2% number. As I mentioned in Analytics Toolbox #1, it is not uncommon for highly skilled players to outscore their expected goals, and the same is true for highly skilled lines.
Alright, that’s enough on the forwards — let’s look at defense! All pairings with more than 150 minutes together at 5v5 are shown in the following table, again taken from MoneyPuck.com.
The first thing that immediately jumps out to me is this: Adam Pelech is a fantastic defenceman. The common denominator on the Islanders’ most dominant two pairings this season is Pelech. The Pelech-Mayfield pair is 17th in the league (among pairs who have played at least 150 minutes together) in ‘xGoals Against Per 60 Minutes’ — that is elite defense. If you raise the ice time requirement to a more reasonable minimum of 300 minutes, they rank 7th.
Assuming they can lock him up at reasonable cap hit, the Islanders need to re-sign Mayfield. I know Pelech-Pulock has historically been ‘the pair’ for the Islanders, but the success of the Pelech-Mayfield pairing gives Barry Trotz the opportunity to spread the wealth by splitting up Pelech and Pulock.
That’s all well and good – it looks like Trotz is doing a fine job at deploying most of his defencemen. The bottom of the table is where things start to get a bit dicey. Barry Trotz and Lou Lamoriello need to fire Zdeno Chara into the moon. Ok, maybe that’s a bit dramatic, but they should have at least scratched him more often this season, and he certainly should not be re-signed.
The table below shows the expected goals for percentages (xGF%) of Chara’s three most common partners this year when they are playing with him and without him.
|
|
xGF% with Chara | xGF% without Chara |
|
Noah Dobson |
48.1 | 48.1 |
| Ryan Pulock | 43.3 |
49.9 |
| Sebastian Aho | 31.8 |
48.1 |
Those numbers don’t paint a pretty picture. Chara considerably drags down Pulock and absolutely craters Aho. An expected goals for percentage of 31.8% is terrible. I’ll caveat that by noting Chara and Aho have only played about 60 minutes together at 5-on-5, so it’s possible the numbers would improve if they had more opportunity to play together. Personally, I wouldn’t risk it – at this point in his career, I believe the most suitable spot for Chara is the press box.
The biggest victim here is Noah Dobson, Chara’s most frequent defence partner. He carries their pair – Chara’s expected goals for percentage drops to a measly 40.5% away from Dobson, 8 percentage points lower than when they are together. Dobson is having an amazing year, but we can only imagine about how much better it could have been if he didn’t have to spend over half his time volunteering at the old folk’s home.
With the season all but lost – and the Mayfield injury to complicate things – it’s not as easy to justify taking Chara out of the lineup now as it would have been earlier in the season. Regardless, it would have been prudent and perhaps advantageous to give players like Robin Salo and Grant Hutton longer looks in the first half of the season. To be fair, Lamoriello bears that responsibility at least as much as Trotz does.
Closing Remarks
Barry Trotz appears to prefer experience over results and stubbornly sticks to his preferred line combinations. These qualities served him well in the past, especially in deep playoff runs where the rules are thrown out the window and the ‘veteran presence’ has a more tangible effect. Unfortunately, I don’t think that style of coaching is what the Islanders needed this year.
For the majority of this season, the Islanders’ young players were not put into positions to succeed. That’s on Barry Trotz. He also put way to much faith in veterans that just aren’t what they used to be. For the Islanders’ sake, I hope he reflects on some of these decisions this offseason. He is an excellent coach with an excellent defensive system, and the Islanders’ can return to being a competitive team next year if he is willing to modernize his player deployment strategy.
I hope you enjoyed our discussion of on-ice metrics today! Stay tuned for Analytics Toolbox #3, where we will discuss a divisive statistic in the hockey analytics community: PDO!
