The thing with Karlsson is that when it comes to xGA Vs. GA, this wasn't his worst year. It wasn't even close.
Just a disclaimer for the rest of this, I am using Natural Stat Trick's expected goal model. The reason I am doing this is because NST is by far the smoothest site and it easily allows me to look at players' teams with and without them.
Basically, here's how I looked at things. If a defenseman can impact goals against in a manner that does not impact expected goals against, then that should show in comparison to his team when he is not the ice. You pretty much have to control for the team or else the stat is worthless. You could put a beer leaguer in front of Henrik Lundqvist and he would probably allow fewer goals than expected. That wouldn't mean that he is a good defender or that he does something good in the defensive end that isn't quantified by expected goal stats. It just means he plays in front of a good goaltender and if that were the case, you would expect it to show in the numbers that his goaltender posts when he is off the ice.
For example, if Karlsson really does lead to more goals against than expected goals and his team has average goaltending, maybe his team would allow 11 goals for every 10 expected goals he allows, and 10 goals for every 10 expected goals that they allow without him. If the team just has good goaltending or faces bad shooters - say they allow 10 actual goals for every 10 expected goals with Karlsson on the ice, and 9 actual goals for every 10 expected goals - then we can still say Karlsson causes a problem.
In both of these examples, I am calculating his team-relative xGA/GA. So, for the first example of his team allowing 11 goals for every 10 expected goals with him and 10/10 without him, here is how you would calculate it:
| Team | GA | xGA | Impact | Impact Percentage |
| Team with Karlsson | 11 | 10 | (11-10)=1 | ((11/10)-1)*100=10% |
| Team without Karlsson | 10 | 10 | (10-0)=0 | ((10/10)-1)*100=0% |
| Karlsson Relative Impact | | | | 10%-0%=10% |
[TBODY]
[/TBODY]
Basically, you calculate the percentage of extra goals they allow above expected, and compare them to their team without them. It's pretty straight forward stuff but it's kind of hard to explain.
Kind of like stats like CA/60Rel and isolated defensive impact stats, being in the negative is a good thing because you allow less. In this hypothetical case, Karlsson has a 10% positive impact on the goals above expected that his team allows when he is on the ice.
Just to explain things one more time, I'm going to use a slightly different hypothetical example in which Karlsson does a good job of suppressing goals against in comparison to actual goals.
| Team | GA | xGA | Impact | Impact Percentage |
| Team with Karlsson | 9 | 10 | (9-10)=-1 | ((9/10)-1)*100=-10% |
| Team without Karlsson | 10 | 10 | (10-0)=0 | ((10/10)-1)*100=0% |
| Karlsson Relative Impact | | | | 10%-0%=-10% |
[TBODY]
[/TBODY]
Now, are you ready for the real Karlsson?
2018-2019:
| Player | GA | xGA | Impact | Impact Percentage |
| Erik Karlsson | 72 | 56.38 | 15.62 | 27.70% |
| Erik Karlsson Team Without | 189 | 164.87 | 24.13 | 14.64% |
| Karlsson Relative Impact | | | | 13.07% |
[TBODY]
[/TBODY]
That's this season. It's worth noting before saying anything else that the reduction in goaltending pads this season has most expected goal models to underestimate the number of expected goals that a team allows. But you're right, everything was going in. His team allowed 15.62 goals more than expected. They allowed 27.70% more goals than expected. That is ridiculous. No matter how much worse than expected those shots actually were, that is still a beer league performance.
However, take a look at the team without him. Yeah, they were still awful. They allowed 24.13 goals more than expected; actually more goals above expected than Karlsson allowed. However, the
percentage of goals above expected that they allowed without him is actually about half as much as the percentage that they allowed with him. They faced far more expected goals and expected goals without him so obviously the sum of goals allowed above expected is higher without him. But make no mistake here -
Karlsson's presence clearly leads to an increase in goals allowed above expected. They allow 14.64% more goals than expected with him and 27.70% more goals than expected without him. He makes terrible goaltenders worst.
I think you're smart and capable of comprehending all of this, so I'm hoping you get where I've gone so far. I'm just typing this message up without any response so I don't know exactly where you're at in terms of comprehension here but I'm making the assumption that you know what is going on here. I also haven't explained this stuff to anybody except for my mom so I'm not sure how difficult it is to comprehend. Even if you aren't 100% caught up, I suggest that you keep reading and I think that things will actually become much more clear.
This could have just been a bad season, right? I certainly agree that by the eye test, it appears that Karlsson was unlucky all year. When I went into looking at these numbers, I realized that this was something that would heavily fluctuate just based on luck, even if guys didn't individually impact this.
Let's look at other years for Karlsson.
2017-2018:
| Player | GA | xGA | Impact | Impact Percentage |
| Erik Karlsson | 121 | 85.28 | 35.72 | 41.89% |
| Erik Karlsson Team Without | 178 | 171.42 | 6.58 | 3.84% |
| Karlsson Relative Impact | | | | 38.05% |
[TBODY]
[/TBODY]
Yikes. Sure, Craig Anderson sucked and Ottawa was a tire fire, but they still almost matched their expected goals against without Karlsson on the ice. Why did their actual goals against blow their expected goals out of the water without him? Maybe he was just unlucky again.
2016-2017:
| Player | GA | xGA | Impact | Impact Percentage |
| Erik Karlsson | 85 | 81.54 | 3.46 | 4.24 |
| Erik Karlsson Team Without | 133 | 152.85 | -19.85 | -12.99 |
| Karlsson Relative Impact | | | | 17.23% |
[TBODY]
[/TBODY]
As a whole, the Sens had very good goaltending this year, stopping over 16 goals more than expected. That is a big part of why they made the ECF. But with Karlsson on the ice, they still allowed more goals than expected. Why?
2015-2016:
| Player | GA | xGA | Goaltending Impact | Impact Percentage |
| Erik Karlsson | 115 | 97.08 | 17.92 | 18.46 |
| Erik Karlsson Team Without | 140 | 169.03 | -29.03 | -17.17 |
| Karlsson Relative Impact | | | | 35.63% |
[TBODY]
[/TBODY]
This is Martin Jones level goaltending with Karlsson on the ice and elite goaltending without him on the ice. What's going on here?
2014-2015:
| Player | GA | xGA | Goaltending Impact | Impact Percentage |
| Erik Karlsson | 89 | 79.27 | 9.73 | 12.27 |
| Erik Karlsson Team Without | 136 | 175.15 | -39.15 | -22.35 |
| Karlsson Relative Impact | | | | 34.63% |
[TBODY]
[/TBODY]
This time, we've got bad NHL level goaltending with Karlsson on the ice, and true Vezina level goaltending without Karlsson on the ice. After looking at this, I was seriously puzzled. Convinced, in fact, that Erik Karlsson did definitely do something on the ice that drove his actual goals above expected. But I had to look at other defensemen. Here are some of the guys that I compared Karlsson to:
| Player | 2014-2015 | 2015-2016 | 2016-2017 | 2017-2018 | 2018-2019 | 5-Year Sum of Team Relative Impact | Average Team Relative Impact |
| Erik Karlsson | 34.63 | 35.63 | 17.23 | 38.05 | 13.07 | 138.61 | 27.72% |
| Danny Dekeyser | 22.45 | 19.24 | 45.34 | 13.9 | 15.92 | 116.85 | 23.37% |
| Braydon Coburn | 44.79 | 15.15 | 6.96 | 5.13 | 41.04 | 113.07 | 22.61% |
| Morgan Rielly | 10.41 | 30.84 | 38.56 | 26.61 | -0.19 | 106.23 | 21.25% |
| Marc-Edouard Vlasic | 12.88 | 39.58 | 17.64 | 4.64 | 4.66 | 79.4 | 15.88% |
| Justin Braun | 14.71 | 25.05 | 14.89 | 2.21 | 5.38 | 62.24 | 12.45% |
| Brent Burns | 25.64 | 18.37 | -15.4 | 12.07 | -10.05 | 30.63 | 6.13% |
| Brenden Dillon | 29.97 | -36.5 | 0.2 | 31.69 | -8.37 | 16.99 | 3.4% |
[TBODY]
[/TBODY]
I looked at 51 players total; all defensemen who played some time in each of the past 5 seasons. Karlsson's average team-relative impact of goals against over expected goals against was by far the highest of any defenseman on this list. I didn't show them all because I didn't want to overwhelm you with data. The reason that I included DeKeyser, Coburn, and Rielly, is because they were the next 3 defensemen behind Karlsson, and the only 3 to have a higher average team-relative impact than him. The rest were the Sharks defensemen that I included because we know them best and I find that it's always best to look at what you know best when doing this type of analysis.
One thing to note is that their average team relative impact on expected goals was all positive. This was the case for 48 out of the 51 defensemen that I studied I am still trying to look into the nuances of why this is, but my initial conclusion is that most regular NHL defensemen will have a higher team-relative xGA/GA simply because they are playing against NHL level shooters. You bring in a replacement level defender and play him for only 40 games per season, he's only going to play against other team's 4th lines who are generally full of replacement level players that can't shoot. Meanwhile, a full-time NHLer, even one like Brenden Dillon who generally plays against bottom-6ers, is still playing more against full-time NHLers. However, I'm not sold on this explanation yet. I ran a regression and couldn't find a correlation between TOI% QoC and team-relative goals against above expected. That may change as I add more replacement level defensemen with a discernible difference in their TOI% QoC but it hasn't yet. It's possible that this has something to do with which positions play special teams and 6V5 situations - like, maybe this stat will always hurt defensemen because they play more on the PK where xGA tends to exceed GA - I don't know. My biggest fear is that the data is somehow faulty but I don't believe that is the case.
But, let's get back on topic and look at the rest of the Sharks and especially Karlsson. Not only does Karlsson find himself in the mid-30s in 3 out of 5 seasons, he is in the teens in every season. Brent Burns' aggregate is still positive, but has two seasons where he was double digits in the negative. Just in case you're still not 100% on this, let's look at those two seasons in particular:
2018-2019:
| Player | GA | xGA | Impact | Impact Percentage |
| Brent Burns | 106 | 94.47 | 11.53 | 12.20 |
| Brent Burns Team Without | 155 | 126.78 | 28.22 | 22.26 |
| Burns Relative Impact | | | | -10.05% |
[TBODY]
[/TBODY]
As you can see, Burns in 2018-2019 played in front of two terrible goaltenders. The same two goalies as Karlsson. When he was out there, they 12.2% more goals than expected. That's horrible. But without him? They allowed 22.26% more goals than expected. Did he really do a great job of suppressing goals that didn't translate to expected goals? I do believe that Burns was better defensively last year and in his Norris year than the other 3, but I'm not convinced that this is anything more than blind luck.
2016-2017:
| Player | GA | xGA | Impact | Impact Percentage |
| Brent Burns | 81 | 95.61 | -14.61 | -15.28 |
| Brent Burns Team Without | 121 | 120.86 | 0.14 | 0.12 |
| Burns Relative Impact | | | | -15.40% |
[TBODY]
[/TBODY]
This season, the Sharks actually had slightly good goaltending and they pretty much got all of it when Burns was on the ice. It was dead on average without him.
Before I did this study, I assumed that Burns would be the goalie killer. I thought he was one guy whose turnovers and defensive gaffes frequently led to goals against in a manner that xGF simply could not quantify. However, after looking at this, he's actually not bad. In fact, he ranks 36th out of the 51 defensemen that I looked at, and his average team-relative impact percentage of 6.13% is well below the average 5-year average of 10%.
Looking at the other Sharks, they tend to fluctuate from year to year. Vlasic and Braun never get good team-relative goaltending but that's probably because they have almost always, over this time frame, taken on the bulk of the toughest forward matchups which include the best shooters. If you look at the team-relative impact percentage of the 4 full-time Sharks defensemen, the player with the highest team-relative impact percentage is the one who faces the top competition, and the team-relative impact percentage drops as competition drops. Dillon faces the easiest competition, and he also gets the best goaltending relative to his teammates. It's also worth noting regards to competition, that Erik Karlsson's competition is not generally high. In fact, his TOI% QoC over this time frame - 31.81% - is less than that of "sheltered Burns" who clocks in at 32.24%, and Vlasic who clocks in at 32.85%. So, I'm skeptical that competition is killing Karlsson. Maybe it's because he plays so many minutes that he faces top lines and 2nd lines, and his team without him plays against 3rd and 4th liners? However, Drew Doughty and Ryan Suter play a higher percentage of their team's minutes and still don't get crushed in team-relative impact on xGA/GA like Karlsson does.
There's a lot more that I could write but I just wrote way too much as it is. I'm curious to see what you think of all of this. I haven't ran a regression yet to see if there is year-to-year repeatability on a season-to-season level for each individual defensemen. My initial guess from looking at these is that there is not but I'm not sure. It's also worth saying that I am still a huge fan of Karlsson and would re-sign him at league maximum term and AAV if it was between that and keeping him. His ability to generate xGF, drive his actual GF above xGF, and suppress xGA makes him a massive net positive even despite this. But this is a legitimate concern.
