Islanders Goaltender Ilya Sorokin & The Analytics Debate

Ilya Sorokin is an elite goaltender, regardless of which analytics are used.

With NHL teams on break for the 4 Nations Face-Off tournament, fans have been spending more time on social media.

For the New York Islanders, talk has been focused on the future of pending free agent Brock Nelson, but goaltender Ilya Sorokin has also gotten some attention.

The Sorokin debate has centered around private versus public data models, with private typically shedding a more positive light on Sorokin's play, while public tends to be more critical:

[gallery ids="53"]

With advancements in data tracking technology and player models, countless routes can be taken to analyze player performance, including the eye test, traditional stats (GAA and SV%), advanced stats (goals saved above expected (GSAx), high-danger save percentage, etc), and now algorithms that combine all of this data into a single value.

Today, a wide range of evaluative tools are available, and results can differ significantly, especially for a position as unpredictable as goaltending, where the margin for variation is even greater.

This is why it’s crucial to remember that models are just that -- models!

Whether in economics or sports, models are not designed to be the end-all, be-all.

One of the most prominent public websites in the NHL analytics world is MoneyPuck.com.

The website's model predicts the most important factors for winning games, but it still has flaws.

GSAx, which measures how many goals a goaltender has saved relative to the total of goals they were "expected" to save, uses expected goals (xG), one of the model's most prominent stats.

According to MoneyPuck, Sorokin has 7.9 GSAx this season, meaning he has saved 7.9 goals above expected.

This metric analyzes the chance of a goal being scored from the shot's location on the ice, meaning a goal like this had just a 7.9% chance of getting past Sorokin:

Other models -- like Natural Stat Trick -- gave Ehlers 0.10 xG on this goal, already showing there are discrepancies in this very xG metric.

The issue here is pretty clear, as these metrics fail to account for the dozens of additional components that factor into each play.

Was the defenseman out of position? Was there a miscommunication? Was the shot sent straight into the goaltender's glove, or was it a perfectly placed shot that went off the crossbar and in? Was this shot following a rebound or cross-crease pass? Was it on a breakaway or odd-man rush? 

You get the point.

In addition to the surplus of private and public models, players can also be evaluated with venue adjustment, which accounts for home ice advantage and the leading or trailing score effects.

This adjustment rewards those who thrive under the bright lights while lowering the value of goals scored in garbage time.

So, which model is right? The answer is simple:

None.

These models exist to complement the eye test, not be the objective truth behind a player's performance.

A prime example is future Hall of Famer Alexander Ovechkin and his goal-scoring ability.

According to MoneyPuck, Ovechkin has scored fewer goals than he was "expected" to score just once in his 17-season career.

Why is that? The model doesn't account for his unworldly shooting talent.

Sure, the website has an additional metric that accounts for his shooting talent, but even those values fluctuate and are not used in GSAx.

For example, his most recent goal on Feb. 6 was a perfectly placed shot following a cross-ice pass, yet it had just a 7.8% chance of going in, according to MoneyPuck:

So, does this mean the model should be thrown out the window?

Of course not, but all models should be utilized in moderation. And it's best to compare a player's season-to-season performance rather than player-to-player.

Utilizing analytics this way helps reinforce ideas like Anders Lee’s adverse luck in 2023-24 when he recorded only 13 goals despite generating 25.3 xG at 5v5. 

Meanwhile, in 2024-25, with a comparable xG rate, he has already scored 16 goals on just 15.5 xG at 5v5.

For goaltenders, this same method should be applied.

One of the biggest success stories this season has been Los Angeles Kings netminder Darcy Kuemper, who has gone 16-6-6 with 10.2 GSAx.

Yet, it would be unfair to say he has played better than Sorokin inherently.

Kuemper has benefited from playing behind a Kings' defense that has allowed just 2.16 expected goals against per 60 minutes (the best in hockey) and 9.37 high-danger shot attempts against per 60 minutes (third best in hockey) at 5v5 this season.

Kuemper has faced just 2.14 expected goals against per 60 minutes and 9.09 high-danger shot attempts against per 60 minutes at 5v5. If these metrics were held across all Kings' games this season, they would rank as the league’s best defense in both categories.

On the other hand, Sorokin has faced 2.35 xG against per 60 minutes and 10.6 high-danger shot attempts per 60 minutes at 5v5, each ranking him around league average.

There is no correct method to weigh advanced metrics with the eye test, as countless variables go into each play on the ice, which is why it's important to diversify and understand which data are incorporated when analyzing players. 

Stay updated with the most interesting Islanders stories, analysis, breaking news and more! Tap the star to add us to your favorites on Google News to never miss a story.