After digging into Lane Lambert’s impact on Kraken goal scoring in my last piece, I couldn’t shake the feeling that there was still more research to look at or areas to explore. Goal scoring tells part of the story, but it doesn’t tell us how teams are generating their chances. In this follow-up piece, I’m zooming in on shot quality: high‑danger, medium‑danger, and low‑danger attempts across Lambert’s coaching tenures, plus a look at expected goals to see whether the underlying process matches the results. If the first article raised good questions, this one aims to push a little deeper into the answers.
In that first piece, I looked at whether Lambert’s system might be suppressing Kraken goal scoring. The short version: the picture wasn’t nearly as straightforward as the raw totals suggested. Yes, Seattle’s scoring dipped in Lambert’s first season, and yes, several players saw their production slide. But when I compared his Kraken tenure to his time with the Islanders, and to the coaches who came before and after him, the results were mixed.
Lambert’s Islanders actually scored more than Barry Trotz’s and nearly as much as Patrick Roy’s, and individual player trends didn’t point to a consistent pattern either. The broader takeaway was that Seattle’s scoring struggles likely stem from a blend of roster construction, usage, and variance rather than a system that inherently throttles offense. That left one big question on the table: if the goals dried up, what was happening underneath the surface in terms of shot quality and expected goals?
Shot quality and expected goals primer
Before we get too deep into the weeds, it’s worth laying out why shot quality matters, and offering a quick primer for anyone who doesn’t spend time dabbling in the dark arts of hockey analytics. Not all shots are created equal. A wrist shot from the point with no traffic is not the same as a backdoor tap‑in rebound, and analytics try to capture that difference by categorizing attempts into high‑danger, medium‑danger, and low‑danger areas. High‑danger chances come from the slot and net‑front, where shooting percentages spike. Medium‑danger chances live in the middle of the ice, and low‑danger attempts are typically perimeter shots with a low probability of beating an NHL goalie.
Expected‑goals (xG) models take all of this into account: location, shot type, and timing relative to other events, to estimate how likely a shot was to become a goal. For this exercise, all shot‑quality and xG data comes from NaturalStatTrick.com, one of the most widely used public analytics sites in hockey. If the first article focused on what happened to Kraken scoring under Lambert, this section is about understanding how those chances were generated in the first place. To illustrate the value of high‑, medium‑, and low‑danger shots, take a look at the shooting percentages across those three categories.
Shortcomings of public models
It’s also worth acknowledging the limitations of the public shot‑quality and expected‑goals models used here. Sites like NaturalStatTrick do incredible work with the data they have, but that data is inherently limited. Public models only track a handful of events: shot attempts, their location on the ice, and timing relative to other recorded events.
What they don’t capture is often just as important: passing sequences, pre‑shot movement, traffic in front of the net, screens, odd‑man rushes, defensive breakdowns, or where the other nine skaters are positioned when the shot occurs. All of those factors meaningfully influence scoring probability, but they’re invisible to public datasets. So while these models give us a useful directional understanding of shot quality, they’re not a perfect representation of what happens on the ice.
High‑Danger chance rate per game
Let’s start with high‑danger chances per game for both the Islanders and Kraken across coaching tenures.
The trends look similar to the goal‑scoring analysis from last week’s Data Dump. Across both teams, the tenure with the most high‑danger chances per game was Lambert’s time with the Islanders, and the tenure with the fewest was Lambert’s time (so far) with the Kraken. That contrast is interesting and a reminder that context matters.
Medium‑danger chance rate per game
Now let’s take a look at medium‑danger chances per game.
The pattern is similar to high‑danger chances, with one notable difference: the drop from Bylsma to Lambert is smaller here than it was in the high‑danger category.
Because the impact is so minimal, I’m skipping a deep dive into low‑danger chances.
Putting it together with expected goals
Continuing with NaturalStatTrick data, here’s the expected goals per game by coaching tenure.
Seeing Lambert’s Kraken tenure produce the highest expected goals per game is a little surprising, which makes me theorize that the 2025–26 team was unlucky, or that the prior four seasons were a little lucky. The truth is probably somewhere in the middle.
For the heck of it, let’s also look at average goals per game by season versus expected goals per game by season.
Truth be told, it’s not as simple as “lucky” or “unlucky,” but the public models do suggest the Kraken deserved more goals than they scored under Lambert last season.
What we learned
I am not sure we gained any additional insight with this additional analysis. What we did get was a baseline: Lambert’s Kraken generated enough expected goals to deserve more than they got, but they also produced the fewest high-danger chances in the league. That’s a tough combo. Call it unlucky, call it unsustainable, call it whatever you want, but if the Kraken want to score more this season, they’ll need to spend a lot less time on the perimeter and a lot more time in the places where goals are actually scored.