Slow Starts, Fast Adjustments: The Islanders Game Script

It’s been a pretty good week for the Islanders. Two wins over the Penguins, one over the Devils, and a tough overtime loss against Montreal leave the Isles in first place in the Metropolitan Division with 18 points in 14 games. One of the more interesting aspects of this streak, and the season at large, is how Islander games have generally progressed.

Typically utilized from a football perspective, Game Script essentially describes how a game unfolds. Mostly tied to scoring margin, Game Script can somewhat portray how teams will play in certain situations over the course of a game. There’s some level of a repeatability factor to it. In other words, teams generally play a certain way in certain situations. For example, hockey analysts sometimes refer to “score effects,” with the theory being that teams will push play more when they are down a goal in the third period. It’s how score-adjusted possession metrics were ultimately introduced into the fold.

Cool, Why Is This Relevant?

It turns out there’s actually some interesting Islander data hidden at a period-by-period level through the first 14 games of the season. In other words, we can actually apply the general thinking of Game Script to Islander games.

Using score-adjusted metrics from Natural Stat Trick, I exported data from each of the team’s games this season and aggregated that data into period level metrics. Do the Islanders have more success in the first period versus the second period? Or the third versus the first? Theoretically, there’d be a lot of noise in a question like that.

But, it turns out there’s actually quite a bit of interesting information.

But First, Let’s Show the Noise

Whoa! There’s a lot happening here. What this shows is the outcome over time by period for each of the five key attempt types. There’s not a ton to take away from this, but two things we can see off the top. First, as the sample size of attempts (as more games are played) gets bigger the line charts are starting to stabilize. It’s expected that continues moving forward.

Secondly, the main contextual takeaway is that the Islanders’ first-period line graph is situated clearly below the second and third periods. Since the Y-Axis describes the percent share of attempts, this immediately tells us that the Isles are generally getting less of the attempts in the first period versus the other two over the course of the season.

Some Improvement?

This next chart is similar to the first, but is a rolling average of the team’s last 10 games. Specifically, how this works is that we take the results of the previous game (and the nine before it), and add those as an entry onto the chart. This is a way to stabilize a lot of the noise, plus create an understanding of how the team is trending at the moment.

It turns out that the Isles are actually improving in their first periods as of late. We can see this in all metrics, especially high danger chances, as the line graph trends upwards towards the latter two periods. Interestingly, both the second and third period graphs still remain fairly stable with standard ebbs and flows. This gives a clearer context than the previous chart allowing us to understanding some recent trends.

The Barry Trotz Effect

So, okay, we have a general baseline now of how the Islanders play in each period. Now. let’s talk about the data itself. As mentioned, all of the following data is 5v5, score adjusted, and comes from Natural Stat Trick. Additionally, I pulled in all shot attempts, unblocked shot attempts, shots on net, scoring chances, high danger chances, and goals, aggregating them by period. The end results immediately stuck out to me.

Not only did the results stick out, they are clear. The Islanders have generally been an atrocious first period team (though as noted before, are seeing slight improvement in recent games). Their first period unblocked attempts, shots on net, scoring chance, and high danger attempt share are all under 40% for the season. It’s unclear specifically why that is the case, but it’s evident.

Things take a clear uptick in the second period, and then again in the third period. In fact, the Islanders are actually over 50% in third periods in shots on net and are pretty close in high danger chances. Compare that to their first period results, which paints a pretty distinct picture of progression within the game.

Interestingly, this even ties back to their goal share. The Islanders are above 50% in every period as it relates to goals, which is a function of many things (but especially continued stellar play from Thomas Greiss and Robin Lehner). But, their first period is their worst goal share period as well, with each period getting progressively better.

Even though the Isles are likely due to some regression in terms of goals against (unless Greiss/Lehner continue to play out of their minds), this does jive with the general idea that more meaningful shot attempts will lead to more goals in the long run. In other words, as the Isles get more of the scoring chance and high danger chance share in latter parts of the game, their goals for percentage goes up along with it.

What Does This All Mean?

At this point, we have a pretty defined typical game flow for the Isles. Start slow, improve in the second, improve again in the third. That’s not necessarily a bad thing, though goaltending has been a key cog in ensuring the Isles aren’t falling behind in their slow starts. To expect that to continue puts a lot of additive pressure on the goalies, so it’s fair to pinpoint the team’s starts as a clear area of improvement going forward.

However, it also calls out the coaching ability of Trotz and his staff. The Isles are still learning new structure, and it’s fair to assume there will be some inconsistency in play through the first quarter of the season under a new coach and with a new system. Frankly, the ability to improve in the midst of a game - likely through systemic and/or tactical adjustments - speaks positive volumes about Trotz’s prowess in his role.

It’s also important to mention that these metrics are not necessarily predictive, but simply a report of how the team has performed in certain situations to date. There are a ton of dynamics at play here: roster churn, system adjustments, player confidence, team matchup, game venue (home vs. away), and more. So we shouldn’t necessarily assume that both the improvement progression and/or their poor first periods will continue.

The reason this is important is because relying on that progression ignores the point of view from the opposition. Other teams are looking to adjust to the Islanders, too. So it is important their first periods improve in the long run, or else it sets up a quite fragile script, which relies upon early goaltending and continual coaching adjustments that lead to improved play with very little margin of error.

But for now, the script has been written.

Thanks to Eric, a noted Radek Martinek enthusiast, for the hypothesis and idea surrounding this piece.

All data above is from Natural Stat Trick and is score-adjusted at 5v5 play.