One late-November afternoon, I was perusing HockeyViz–as I am wont to do. I almost always come away having learned something new about a player, a team, or how NHL hockey is played, and that is what happened on that day.

Parsing through Joey Daccord’s data, I found that in virtually every gameplay situation, the Seattle Kraken generate better shot quality offensively and significantly suppress opponent shot quality defensively when Daccord is on the ice, relative to when another goalie is playing.

This type of finding is not atypical for a skater. We would say that skater plays a “two-way game,” showing effective defense to prevent dangerous shots against and offense to generate those types of shots. HockeyViz credits Jared McCann, for example, for value-generating play driving in all situations–even-strength offense, even-strength defense, power-play offense, and penalty-kill defense. It’s easy enough to understand how this would be the case.

But a goalie’s job is to save shots on goal. We have conventionally measured the goalie’s worth via the save percentage stat, i.e., what percentage of shots on goal the netminder turns away. Goalies who save a higher percentage of the shots they face are the “good” ones.

More recently, we in the hockey community have increasingly utilized metrics developed by various outlets commonly called “goals saved above expected.” These metrics incorporate data on shot location and shot type faced by a goalie to estimate how many goals an average goaltender would be expected to concede based on the shot quality against. This expected outcome can then be compared against the goalie’s actual performance. Goalies who save more shots than expected are the “good” ones.

Either way, the focus is always on what a goalie does with shots faced. This is because a goalie can’t be asked to prevent those shots in the first place. That’s the skaters’ role, and it’s certainly not the netminder’s job to create offense on the opposing goalie. Right? Or is it?

The finding that the Kraken create better offensive chances and suppress chances against more effectively when Daccord is in goal has held consistent over the last few weeks since I began tracking it. 

For example, let’s look at shots the Kraken have given up at five-on-five in the first 39 games of the season (through the Jan. 4 game against Ottawa). Areas of the ice where the Kraken give up fewer shots than average (good) are represented in progressively darker shades of blue. Represented in red are areas where the Kraken give up more shots than average (maybe bad, but shots from the perimeter can be ok).

In Joey Daccord’s five-on-five time on the ice through Jan. 4 (1119 minutes of game play), the Kraken allowed opponents shots valued at 15 percent lower than league average according to HockeyViz. This is very good. When Daccord is not on the ice five-on-five (774 minutes of game play), the Kraken have allowed shots valued at nine percent lower than league average. This is still good, but significantly less so. 

Now, let’s look at the shots the Kraken have generated five-on-five. Areas where the team generates more shots than average are in red (good), whereas areas of fewer than average shots are in blue (bad).

Again, the Kraken results have been better in Daccord’s minutes. The team has generated better quality chances, valued at just two percent below average when Daccord is in net, versus eight percent below average when Daccord is not in goal.

If we dig into the numbers in other personnel situations–which we will–we find a similar disparity. 

Is it possible that there is a causal relationship here? Could Daccord really be responsible for a material change in the gameplay around him? Inspired by a number of related questions about randomness, sample size, causation, how we use advanced shot metrics to measure the value of goalies, and how Joey Daccord plays the position, I set out to investigate. 

I came up with six potential explanations that I believed could explain some or all of the observed shot quality disparities.

1. Quality of competition
2. Quality of teammates
3. Score effects
4. Change in Kraken skater performance independent of Daccord’s play
5. Randomness/small sample size/noise
6. Joey Daccord is directly affecting Kraken shot quality

After working through these possibilities, I have settled on a working theory that I will outline here and parse in greater depth in a follow-up article at a later date. Let’s dive in.

Quality of competition

If Daccord has faced a weaker schedule of offensive and defensive competition than the other goaltenders, this could explain some or all of the observed shot quality disparity. 

Consider a scenario where Philipp Grubauer has started one game against the 2023-24 Vegas Golden Knights, and Daccord has started one game against the 2023-24 Chicago Blackhawks. Kraken shot quality may be much stronger in Daccord’s start, but this could just be about the strength of the competition.

I used Evolving Hockey’s game log tool to determine the relative strength of opposing offenses and defenses faced by Daccord and by Grubauer/Driedger based on Evolving Hockey’s expected goals models. Where both goaltenders played in a game (e.g. Grubauer played two periods only to be replaced by Daccord for the third period), I was able to apportion the game based on each goalie’s time on ice. 

I found that the opposing offenses faced by Daccord as opposed to Grubauer/Driedger where roughly similar, but the Kraken were far more effective suppressing shot quality in Daccord’s starts. On the other hand, the opposing defenses Daccord has faced have been stronger than those matched up against Grubauer/Driedger. Even so, as we saw above, the Kraken were more effective generating shot quality in Daccord’s starts.

Daccord has faced opposing offenses that have averaged 3.25 expected goals per game for the season. Yet, against the Kraken, those teams have managed to produce just 2.76 expected goals per game, a decline of 15.2 percent below their average production for the year.

Grubauer and Driedger have faced opposing offenses that have averaged 3.30 expected goals per game this season. These offenses have been only mildly more threatening on average (+0.05 xG/game) than the offenses faced by Daccord. Yet, those offenses have managed 3.04 expected goals per game against the Kraken, a decline of just 7.7 percent below their season-average production.

The opposing team defenses in Joey Daccord’s time on ice have conceded just 3.07 expected goals per game on average, whereas the defenses in Grubauer and Driedger’s time have conceded 3.29 expected goals per game. Even so, as detailed above, Seattle’s overall shot quality has been stronger in Daccord’s starts.

In sum, quality of competition does not explain the observed disparity in shot quality when Daccord starts. If anything, the Kraken have faced more difficult opposing defenses in his starts, so one would expect to see lower expected goal production for the Kraken in Daccord’s starts. The opposite has happened.

Quality of teammates

If Daccord has had a healthier and more productive roster of Kraken players during his time on ice, one might expect superior shot quality outcomes. But, here again, Daccord has faced a roughly similar support context as the other goaltenders. 

The Kraken blue line has remained healthy for most of the year. Most of the variation has come in the forward group. 

Grubauer was operating as the lead goaltender until an injury landed him on injured reserve on Dec. 10, 2023. While healthy, he played a majority of his starts without Andre Burkavosky, a top-six forward, and Brandon Tanev, a bottom-six forward. Jaden Schwartz, another top-six player, was placed on injured reserve 11 days before Grubauer, so Grubauer missed Schwartz’s contributions for that time as well. 

For his part, Daccord played the large majority of his starts without both Schwartz and Burakovsky, two projected top-six forwards. Recently, he has started without Pierre-Edouard Bellemare, the team’s fourth-line center–an absence which may be felt on the penalty kill due to Bellemare’s face-off and shot-blocking abilities. 

I don’t view these relative absences as a probable explanation for the observed shot quality disparity.

Score effects

There is a well-documented phenomenon in hockey that trailing teams outshoot their opponents and leading teams get outshot, and the effect becomes more extreme the closer to the end of the game. 

The best research in the area tends to indicate that this is because leading teams “hang back” and play more conservatively, rather than some “improvement” in the compete level or production of their trailing opponents. Micah Blake McCurdy of HockeyViz has done research that suggests expected goal production by a team with a lead declines around 10 to 15 percent, while, on average, the trailing team continues to push at a pace similar to its level when the score was tied. 

As relevant here, the score effect phenomenon suggests that–through no inherent merit–a goalie may see relatively favorable aggregate shot quality for and against the goalie’s team if the goalie spends most time on the ice while the team is trailing. This is because the goalie’s teammates will continue to push for offense through that time, while the opponents ease off and play more conservatively. 

On the flipside, the shot quality for and against a goalie’s team may look less favorable if the goalie’s team spends much of the goalie’s time on ice protecting a lead. The team’s offensive numbers will decline while the opponents continue to push for an equalizing goal.

Here, HockeyViz data indicates that 43 percent of Daccord’s time on ice has been spent protecting a lead, against just 25 percent of his time behind on the scoreboard.  In contrast, the Kraken have been protecting a lead in just 21 percent of the time Grubauer and Driedger have been in goal, against 44 percent of the time trailing.

This suggests that score effects are not the cause of the observed shot quality disparity while Daccord is in goal. To the contrary, if we were to adjust for the impact of score effects–which I have not–the observed shot quality while Daccord is on the ice is significantly more impressive relative to the other goaltenders.

Change in Kraken skater performance

Another potential explanation is improved play by the Kraken skaters while Daccord is in goal unrelated to Joey Daccord himself. In other words, it could be a coincidence that Daccord has been in net when the skaters are performing better.

There is at least superficial appeal to this theory. Joey Daccord has assumed most of the team’s starts in goal since Grubauer landed on the injured list on Dec. 10. And that time has coincided with a long point streak the team has at least partially attributed to a tightened defensive approach.

As reported by Sound Of Hockey’s Darren Brown, Kraken defenseman Vince Dunn described an early-December team meeting as the jumping-off point for this modified mindset. “[We were] really chatting about our team identity, not looking at video and things like that,” Dunn said. “I think since that meeting, we’ve all been a lot more accountable. And we’ve been taking every game and our jobs a lot more personally and making sure that if we work hard and work smart and care for each other out there, then things should fall into place for us.”

Said Kraken coach Dave Hakstol after the Winter Classic, “In order to gain some traction and push the right direction here, we’ve had to do it with a defensive mindset and a real grinding type of mindset. We still have a lot of work left to do, but we’ve… look at it from where we were a couple weeks ago to where we are today. We’ve given ourselves a chance and an opportunity to keep climbing and keep working towards something.”

During the recent winning streak, we have also noted a more aggressive and effective approach by the penalty killing skaters. 

While these factors likely have contributed to recent improvements in suppressing opponent shot quality, I don’t think it fully explains what I have observed in Daccord’s numbers. First, an increased emphasis on defense does not directly explain the observed offensive shot quality gains the team has enjoyed with Daccord in net. 

Second, and more importantly, the timing of the team’s defensive adjustments noted by Dunn and Hakstol does not materially change my observation that the team suppresses shots against at a higher rate with Daccord in goal. Recall that I began working on this analysis because of numbers I was seeing in late November, before the Grubauer injury and recent win streak.

As I calculated above, overall the team has performed 15.2 percent better than “average” in suppressing opponent shot quality with Joey Daccord in goal. If we take out all of Daccord’s starts during the current point streak (since Dec. 12) and consider only his starts before then (up through Dec. 10), the Kraken performed 16.2 percent better than average suppressing opposing shot quality when Daccord was in goal during that time. In other words, Daccord is not simply “benefiting” from a recent run of strong play in front of him.

I don’t want to discount what the players are saying about scheme and mindset adjustments, but these numbers make me think that the recent strong defensive work has more to do with (1) the team entering a softer portion of the schedule, and (2) Daccord starting nearly every game in goal.

Finally, I should note one related potential explanation. It has been suggested that teams may play with greater confidence if a certain goaltender is in the net, which itself can improve the quality of skater play even though the goaltender is not having any type of direct impact. For example, Brown has made this case on The Sound Of Hockey Podcast when discussing Seattle’s run of strong play in front of Martin Jones last year.

I don’t have data to discount this theory necessarily. To the contrary, it is potentially supported by my discussion above that the Kraken have been better at suppressing shots with Daccord in goal throughout the season–even when the team wasn’t winning as much. And I chose Daccord as Seattle’s best “hype man” for a reason.

Still, I have a hard time rationalizing this as “the explanation” given the magnitude of the shot quality variation in all game play situations. Also, the Kraken themselves have not put any weight behind the theory. 

Regardless, if it is true Daccord has an innate ability to inspire confidence and improved play from his teammates, I suppose this trait–like his puck play–is value brought by Daccord himself beyond his “goals saved above expected.”

Randomness, small sample size, or noise

Lastly, it could be that the observed shot quality phenomenon is simply random noise in the data that would disappear in a larger sample. Up front, I’ll concede that this could prove true. That said, four factors suggest to me that the observed shot quality disparity isn’t completely random.

First, the numbers have been fairly consistent throughout the season–and across all contexts. I provided the five-on-five shot charts above. Here is the shot quality disparity on the penalty kill:

The team suppresses shots nine percent better than average on the penalty kill when Daccord is on the ice. When he’s not? The team is 28 percent worse than average with other goalies.

Here is the disparity on the power play:

Again, the team performs better with Daccord in net. The Kraken generate shot quality four percent better than average with Daccord, and 10 percent worse than average without him.

These disparities across all contexts are fairly large as well. This isn’t a case where one or two games could swing the analysis in the other direction.

Second, halfway through the season is when data analysts tend to begin publishing their findings on single-season shot data because the sample size has gotten large enough. With 40 games played, we’re in the right ballpark.

Third, a quick look at other historical goaltenders who have gained a reputation as “good puck handlers” supports the premise that a goalie may have a real, repeatable impact on the shot quality around him. 

For example, Carey Price is often mentioned as the best modern puck-handling goalie. In every single healthy Price season going back to 2016-17–which is as far back as I looked–the Canadiens had a net expected goal differential five-on-five that was stronger when Price was on the ice than when he was off the ice. (And this is potentially even more impressive when you consider that Price was treated as a workhorse starter in Montreal, and his backups likely took easier matchups.)

Fourth, I’ll invoke the “eye test.” My working theory “explaining” the observed shot disparity is built on game observations. I’ll finally get into it in a moment, but I think the tape supports the conclusion that Daccord has a significant impact on shot quality outcomes around him. 

All of that said, I do not want to overpromise on my conclusions here. I think we are years away from understanding whether Daccord’s play style has a significant, sustainable impact on the shot quality around him. The early returns are intriguing though. Let’s turn to that now.

Daccord’s active style is driving play for Seattle

Having found reasons to question the other options, we are left to consider whether something Joey Daccord does has improved shot quality outcomes around him.

My working theory in this regard is that Daccord’s active goaltending style is the reason. This is the obvious distinguishing factor between him and Philipp Grubauer. Daccord is much more aggressive coming out of his net to preempt opposing play and move the puck away from pressure and onto the sticks of his teammates. Daccord pairs this aggressive mindset with high-end puckhandling and passing skills that allow him to execute on plays that other goaltenders wouldn’t dare consider.

Those with a front-row seat to Daccord’s play have offered comments that could support this theory. In November, Kraken defenseman Adam Larsson told Kate Shefte of The Seattle Times that the goaltending play styles of Daccord and Grubauer are “pretty opposite.” “[Daccord]’s out of the net a lot more than [Grubauer]. It’s almost like an extreme in both ways, which is good and bad.”

“[Daccord] has a natural ability [playing the puck],” Hakstol said. “Joey wants to have an effect on every single play, and that’s usually not the way it works. He’s working to continue to pick his spots [playing the puck].”

Consider the following examples.

Here, during the team’s Dec. 14 game against Chicago, Daccord activates from the goal crease to gather a dump-in entry before it reaches the corner. He then scans the ice and delivers a pass off the boards past the first wave of forechecking Blackhawks and onto Devin Shore’s stick.

In contrast, Philipp Grubauer likely would have managed this play more conservatively, either staying in goal, or stopping it for his defensemen to gather and lead the breakout after he had resumed his position in the net. This delays the zone exit sequence and increases the forechecking risk, but mitigates the disaster scenario of a turnover leading to an empty-net goal.

Above are two examples of shoot-ins from Grubauer’s start in San Jose on Mar. 16, 2023, where Grubauer opts to stay in goal rather than stopping the puck and looking to move it onto a Seattle stick. Grubauer protects against the gaffe, but concedes two protracted defensive-zone sequences that he may have been able to break if he were more proactive and efficient with the puck.

The difficulty I had with this “explanation” for the shot quality disparity I was seeing is the sheer magnitude of the effect. I could rationalize active goaltending mitigating a chance here or there, or turning a stretch pass into an offensive shot attempt every few games, but the numbers suggested a much larger impact than I had expected. Could Daccord’s puck play really be responsible for nearly doubling the effectiveness of Seattle’s defense?

I wanted to know, so I started a project of closely watching and accounting this aspect of Daccord’s game for over a month. The results were eye-opening and led me to believe that Daccord’s extreme activity outside the goal could be a far more significant strategic advantage than any public analyst has yet been able to quantify.

I’ll get into that project in my next post, coming soon.