All Purpose Analytics and Extended Stats Discussion

[g00n]

The writeup for this particular model is here:

Wins Above Replacement — High Level Overview

That Tableau isn’t great for selecting by team, unfortunately. I think you can type the three letter team code (i.e. WSH for the Capitals). As far as I can tell that’s the only place the data is available.

Ok, looks like the guy put a lot of work in. It has Kuzy as worst EV WAR over the last 3 years, but next to worst is Ovie, so not sure how useful that model is IRL.

 

[twabby]

Ok, looks like the guy put a lot of work in. It has Kuzy as worst EV WAR over the last 3 years, but next to worst is Ovie, so not sure how useful that model is IRL.

If you’re looking at the EV WAR column then that is only measuring scoring chance generation offensively and suppression defensively without factoring in the quality of the shooter. Ovechkin is actually quite bad in terms of xGF differential over the past 3 seasons, as is Kuznetsov.

But Ovechkin in particular makes up for it because he is such a good shooter. He scores goals from locations that other shooters don’t. As you can see, Ovechkin’s EV Shooting WAR more than makes up for his EV WAR in terms of chance differential. Only Auston Matthews, Leon Draisaitl, and our friend Jakub Vrana have been better shooters over the past 3 years at EV according to this model.

This leads to an interesting discussion about the merits of using GAR vs. xGAR. The models are similar except GAR includes this on-ice shooting component while xGAR does not. GAR is typically used to measure forwards, with the idea being that forwards show marked disparities in their shooting abilities over time. Think Alex Ovechkin vs. Carl Hagelin. Though Hagelin generally is probably a better driver of play, he is such a bad finisher that it should be counted against him.

xGAR, on the other hand, is typically used to measure defensemen. This makes sense because there isn’t nearly as much variability in defensemen’s shooting abilities, at least at EV. Most of a team’s ability to outscore their xG is driven by their forwards.

In particular Justin Schultz looked incredibly bad this year under the xGAR model from Evolving Hockey, but looked quite good under their GAR model, fueled by his EV Shooting GAR. But unless you think his on-ice shooting percentage was driven by him in particular instead of the multitude of good shooting forwards that the Capitals have had for years, I don’t really buy the GAR results. But in Ovechkin's case, yeah I’ll buy that he is one of the best shooters ever.

 

[twabby]

So, if they get half decent goaltending, say 18th overall, they'd move up to what spot as a team overall?

Goaltending is huge. I went through a few teams to see who was projected to be an "average" team in terms of goaltending, and found St. Louis was 17th with Binnington and Husso. Assuming Binnington plays 50 games and Husso plays 32, they project to be 3.1 wins above replacement, compared to 0.4 WAR for the Samsonov/Vanecek duo.

This average goaltending appears to put the Capitals at 99 points, which would put them at around 7th-9th place in the NHL according to this model. Getting average goaltending should be enough to get the team to the playoffs, however they are still quite a bit short of the real contender tier.

Going forward even more, what happens if you replace Kuznetsov with a player like Tomas Hertl or Jack Eichel? Replacing Kuznetsov with Hertl/Eichel adds ~5 standings points to the team's fortunes, which doesn't sound like too much except now you're going from the Pittsburgh/NY Rangers tier into the Boston/Toronto/Vegas tier if they can get average goaltending and they get someone like Eichel. Dumping Schultz for a replacement level on the third pair and promoting Jensen to he second pair gets them ~2 more standings points. Nothing they can realistically do will put them in the Tampa Bay/Colorado tier, but it is possible for this team to become a pretty strong contender with an aggressive move to get a 1C and to get rid of Justin Schultz.

Above is the roster with some changes I mentioned. I put Binnington and Husso as a stand-in for "average goaltending", and inserted Kempny and Fehervary (who are both treated as replacement level players since neither have enough recent TOI to be modeled with any level of confidence).

This roster is only behind Colorado and Tampa Bay in terms of projected standings points. It's a big reason why I am advocating so strongly for an aggressive move to address the center position and to do anything to get rid of Justin Schultz. This team as an opportunity to be very good next year, much better than last year. But without addressing these holes they really aren't that strong.

 

[CapitalsCupReality]

Which is it Caps fans, , we’re struggling to make or we’re missing the playoffs, or we’re projected to earn a lot of standings points?

I side with Twabby. This team isn’t terrible, and it should put up good standings points.

 

[twabby]

Which is it Caps fans, , we’re struggling to make or we’re missing the playoffs, or we’re projected to earn a lot of standings points?

I side with Twabby. This team isn’t terrible, and it should put up good standings points.

They're currently projected to be tied for #4 in the Metro with Carolina, and behind Philadelphia, Pittsburgh, and the NY Rangers. They're only projected to be 2 points ahead of New Jersey, who made a couple of nice additions this offseason and should see some good growth from their young #1 draft picks. For a reason I'm not quite understanding right now, the Capitals are projected to be 5 points ahead of the Islanders.

As it stands they are not really projected to put up "good standings points." They are projected to be incredibly mediocre.

 

[CapitalsCupReality]

They're currently projected to be tied for #4 in the Metro with Carolina, and behind Philadelphia, Pittsburgh, and the NY Rangers. They're only projected to be 2 points ahead of New Jersey, who made a couple of nice additions this offseason and should see some good growth from their young #1 draft picks. For a reason I'm not quite understanding right now, the Capitals are projected to be 5 points ahead of the Islanders.

As it stands they are not really projected to put up "good standings points." They are projected to be incredibly mediocre.

that’s why your stats are shite!

 

[twabby]

that’s why your stats are shite!

Thank you for your contributions to this thread.

 

[ALLCAPSALLTHETIME]

Goaltending is huge. I went through a few teams to see who was projected to be an "average" team in terms of goaltending, and found St. Louis was 17th with Binnington and Husso. Assuming Binnington plays 50 games and Husso plays 32, they project to be 3.1 wins above replacement, compared to 0.4 WAR for the Samsonov/Vanecek duo.

This average goaltending appears to put the Capitals at 99 points, which would put them at around 7th-9th place in the NHL according to this model. Getting average goaltending should be enough to get the team to the playoffs, however they are still quite a bit short of the real contender tier.

Going forward even more, what happens if you replace Kuznetsov with a player like Tomas Hertl or Jack Eichel? Replacing Kuznetsov with Hertl/Eichel adds ~5 standings points to the team's fortunes, which doesn't sound like too much except now you're going from the Pittsburgh/NY Rangers tier into the Boston/Toronto/Vegas tier if they can get average goaltending and they get someone like Eichel. Dumping Schultz for a replacement level on the third pair and promoting Jensen to he second pair gets them ~2 more standings points. Nothing they can realistically do will put them in the Tampa Bay/Colorado tier, but it is possible for this team to become a pretty strong contender with an aggressive move to get a 1C and to get rid of Justin Schultz.

View attachment 459935

Above is the roster with some changes I mentioned. I put Binnington and Husso as a stand-in for "average goaltending", and inserted Kempny and Fehervary (who are both treated as replacement level players since neither have enough recent TOI to be modeled with any level of confidence).

This roster is only behind Colorado and Tampa Bay in terms of projected standings points. It's a big reason why I am advocating so strongly for an aggressive move to address the center position and to do anything to get rid of Justin Schultz. This team as an opportunity to be very good next year, much better than last year. But without addressing these holes they really aren't that strong.

Thanks for the detailed reply. Hope for at least average goaltending then.

 

[Langway]

By the numbers: Grading every NHL team’s contract efficiency, 2021 edition

Caps are at 16th, after placing 23rd last year. Sheary, Sprong & Dowd help based on last year's play but it's interesting to see Carlson near the top in value and Orlov on the lower end on Dom's model. Not sure about that or the likes of McDonagh & Pietrangelo being rated so poorly in projected value. Irwin is included so removing him could bump them up a spot or two. Also worth mentioning that goaltending isn't factored in. The Caps spend less at the position so there's more money for skaters, even if not ideally spent.

 

[twabby]

Talent distribution – Goaltending (Part IV)

Good writeup on the impact, talent distribution, and predictability/repeatability of goaltending.

Summary

• Goaltender performances are almost impossible to predict. It’s difficult so say, if this is due to bad models/data or if goaltending performances just are this random.
• On a seasonal basis the talent distribution is skewed towards the bad goaltenders, indicating that the negative impact of the bad goalies is greater than the positive impact of the good goalies. When looking at career stats the differences are smaller, and almost linearly distributed.
• When you look at total GSAx, there is a pretty good correlation between talent and career length/workload.
• Goaltending is obviously extremely important, but it’s difficult to predict, so overcommiting to one goaltender is a bad idea, unless you’re very confident in his ability. It may be preferable to go with an 1A/1B tandem, since it gives you two chances for a great performance. Very few goalies are good year after year.
• Goaltenders have greater seasonal impact (good or bad) than skaters, but not necessarily greater career impact.

I’m disappointed that Washington wasn’t able to address their need at center but I am very happy that GMBM was able to go extremely cheap on goaltending. I’m not big on Vanecek or Samsonov but the goalie market was wacky this offseason so I’m glad they avoided overpriced veterans who aren’t likely to be any better than the C*pitals duo.

 

[twabby]

A few neat microstats tracked by InStat. No idea how relevant, accurate, or predictive they are, but they do match the eye test for a lot of the players. I'm only going to post a few relevant to the C*pitals, but if you go through JFresh's Twitter account there are more:



If nothing else Nicklas Backstrom is extremely responsible with the puck. Dowd being on there is neat too. Several strong defensive players on the left side of the list.



Honestly not too surprised by Jensen turning the puck over a lot. Matches the eye test where it seems like he sometimes gets himself into hot water, though like a lot of the guys on the list he's usually able to recover with great frequency. A lot of very good defensemen on the list, including Jensen.



A little surprised to see Orlov on this list, but it's not too shocking I guess. Again a lot of very good defensemen on this list.

 

[twabby]

More microstats:



Not surprising to see Backstrom as one of the more frequent passers in the league. Good company to be in.



Orlov the most frequent passer on the Capitals and in the top 20 in the league. Might explain some of why he's so good in transition.



Orlov again near the top of the list, this time in zone entries from defensemen. His strong suit certainly seems to be transition play and gaining the offensive zone, which might explain why his offensive impacts are strong despite not scoring a bunch.

 

[kicksavedave]

Honestly not too surprised by Jensen turning the puck over a lot. Matches the eye test where it seems like he sometimes gets himself into hot water, though like a lot of the guys on the list he's usually able to recover with great frequency. A lot of very good defensemen on the list, including Jensen.

I think the difference between top (19.8/60) and bottom (17.0/60) seems actually pretty small considering they play around 15-20 mins per game - around 9.9 vs 8.5, or 1.5 TO's more per game from best to worst. I also think that D who possess the puck more and try to skate it out, will make more turnovers when compared to D who are quick to dump the puck off their stick out of the zone, chip off the boards, etc. Although I'd like to see the science, because chipping off the boards to alleviate pressure usually ends up on the other teams stick in the neutral zone, does that count as a turnover? If so then this stat is somewhat meaningless, because D men who scramble to lose pucks and chip them out before the forechecker gets there shouldn't be dinged for that when compared to slow D who let the forechecker have the puck in the OZone.

 

[twabby]

I think the difference between top (19.8/60) and bottom (17.0/60) seems actually pretty small considering they play around 15-20 mins per game - around 9.9 vs 8.5, or 1.5 TO's more per game from best to worst. I also think that D who possess the puck more and try to skate it out, will make more turnovers when compared to D who are quick to dump the puck off their stick out of the zone, chip off the boards, etc. Although I'd like to see the science, because chipping off the boards to alleviate pressure usually ends up on the other teams stick in the neutral zone, does that count as a turnover? If so then this stat is somewhat meaningless, because D men who scramble to lose pucks and chip them out before the forechecker gets there shouldn't be dinged for that when compared to slow D who let the forechecker have the puck in the OZone.

It's most to 20th most, not most to least.

I imagine you're right that those who try to skate it out, or at least those who are depended on leading the transition from the defensive zone, will turn the puck over more than those who are content to chip it out. I have no idea what InStat considers a turnover, but I suspect a chip-out or dump-in is not considered a turnover.

I guess for me this is why I haven't taken to microstats yet. Small picture items are neat if you're a coach trying to improve elements of a player's game, but to point to a player and say "he stinks at X microstat" is basically meaningless to me at this point because I have no idea what sort of impact these microstats have on goal differential. At least with other stats like xGF, WAR, etc. there is plenty of work to indicate how these elements impact goal-differential. Microstats like these, as far as I am aware, have not been integrated in any sort of overall player-impact model.

So I think these stats are neat, but ultimately their importance seems to be in question for now.

 

[Midnight Judges]

I think some of those stats could be quite dependent on the system and how the players are deployed and used. For example a puck moving defenseman or offensive defenseman is more likely to have turnovers per 60 than a stay at home defensive player.

 

[Hivemind]

I've usually used Corey Sznadjer's Tableau for microstats. He's also the perfect example to prove all those who think that analytics are for people who "nerds who don't watch the game" wrong. Corey S watches more hockey than just about any of us, as his giant repository of microstats are all manually tracked by him watching hockey.

Tableau Public

It'll be interesting to compare InStat's data against Corey's when I have more time. I know off the top of my head that his data also has Roman Josi blowing away every other defenseman in the league in zone entries, so there's already a data point that matches up. Corey's data often also tracks a stat by multiple measures (raw totals vs success rates or carry-ins vs. dump-ins, etc), so it can often be used to make pretty informed decisions about why things are happening the way they are for a player or team.

With regards to turnovers by D, specifically, these are stats that have been tracked by the NHL for a while. The largest conclusion has always been that "D that hold the puck more turn the puck over more." Raw turnover totals are a result of ice time and time with the puck, and turnovers/60 are a result of time with the puck/60. What would be more interesting now would be InStat starting to directly track individual player possession time, and then mapping turnovers against possession time to get a turnover rate that says something about the riskiness of a player carrying the puck within their system.

Here's JFresh's take on microstats in general.
How Should We Interpret Microstats?

I think JFresh brings up good points, but I also don't have the same faith in all-in-one outcome metrics for hockey that he does (xGF, xGA, WAR-types). Blending multiple stats has always been the way for me, so I do come to the same conclusion, but I still think we have a ways to push forward with our combined metrics and that the advancement of automatically tracked microstats will open up a lot of doors for hockey analytics (the way it has for basketball and soccer). There's a wonderful article from FanGraphs several years back that I can never seem to find when I want to, but it basically was alla bout reconciling the differences between baseball analytics guys and "traditional" coaches. The analytics guy might say that a player's contact rate has fallen, while a batting coach might say he's in a slump because he's late on his swing. Both of those things can be true simultaneously. I think that microstats will be a way to help bridge a similar divide in hockey.

 

[kicksavedave]

All I know about Jensen and microstats is that there was this one time when I called him a p***y and said he would never score a goal ever, and then he scored a goal like 5 minutes later, so we have this kinda connection now.

 

[max21]

All I know about Jensen and microstats is that there was this one time when I called him a p***y and said he would never score a goal ever, and then he scored a goal like 5 minutes later, so we have this kinda connection now.

Dude I remember that! It was against Philly lol that was amazing

 

[twabby]

Here are some of Kuznetsov's stats for this year and last, both boring and fancy (data from Evolving Hockey):

Statistic	2020-21	2021-22
5v5 TOI/Game	13.31	15.82
5v5 points/60	2.09	2.19
5v5 primary points/60	1.76	1.60
5v5 ixG/60	0.53	0.64
5v5 xGF diff. /60	0.32	0.25
PP TOI/Game	2.46	3.62
PP points/60	4.88	5.31
All Sits. GAR/60	0.497	0.541
All Sits. xGAR/60	0.597	0.297
Defensive zone start %	3.05%	3.94%

[TBODY] [/TBODY]

He seems very much like the same player as last year, except now he's getting more TOI at both 5v5 and on the PP, along with getting some PK time due to the extended absence of Backstrom along with sporadic absences of Dowd and Eller.

I am very interested to see what happens to his TOI once all of the centers are back in the lineup.

 

[twabby]

I wanted to add this, because I think this is also a huge contributing factor to why Kuznetsov is perceived so differently this year:

	2020-21	2021-22
QoT Z-Score	-0.53	0.61

[TBODY] [/TBODY]

The Z-Score is measured by isolated xG play-driving throughout the NHL. Last year Kuznetsov had the worst teammates by this metric of any C*pital. This year his teammates are quite a bit better.

 

[g00n]

Or, he's getting more ice time BECAUSE he's playing better, and that's diluting his p/60.

Or, he's playing with better teammates BECAUSE the team is playing better.

There's a reason the coach and front office and Kuzy himself all perceived his performance as sub-par last year, and you didn't have him saying "I'm playing great, I don't know what the problem is, have they checked my fancy stats?"

 

[CapitalsCupReality]

He’s clearly not the same player this year as the past 2.

stop trying to blame anyone but the player for his own performance.

 

[twabby]

Or, he's getting more ice time BECAUSE he's playing better, and that's diluting his p/60.

Or, he's playing with better teammates BECAUSE the team is playing better.

There's a reason the coach and front office and Kuzy himself all perceived his performance as sub-par last year, and you didn't have him saying "I'm playing great, I don't know what the problem is, have they checked my fancy stats?"

Who else would be getting top line minutes if not Kuznetsov? With Backstrom out and Eller/Dowd both missing decent chunks of the season did Laviolette really have any choice but to play Kuznetsov in a number 1 role both at 5v5 and on the PP? With the entire second line out for a large stretch and with an entire line of rookies playing on the 4th line, wouldn't this also force Laviolette to play Kuznetsov more minutes?

Indeed, Kuznetsov played 21:22 in the very first game this season, more than any regular season game he played last season. Since this was before he had established any sort of consistency this season, doesn't this suggest that his increased TOI was, at least initially, due more to circumstances beyond Kuznetsov's control? If Backstrom had been healthy this season do you think he would be getting the same opportunities on the PK, for instance? I doubt it.

I also don't buy that his increased TOI dilutes his points/60. There's no data to suggest that this happens with players around the league who are thrust into bigger roles. If Kuznetsov's minutes were reduced to what he received last year, do you think his points/60 would just increase? Seems unlikely.

I think the team is playing better, particularly Ovechkin, and this certainly impacts the z-score numbers above. But I also think this influences how people are perceiving Kuznetsov's game.

 

[g00n]

Who else would be getting top line minutes if not for Kuznetsov? With Backstrom out and Eller/Dowd both missing decent chunks of the season did Laviolette really have any choice but to play Kuznetsov in a number 1 role both at 5v5 and on the PP? With the entire second line out for a large stretch and with an entire line of rookies playing on the 4th line, wouldn't this also force Laviolette to play Kuznetsov more minutes?

Indeed, Kuznetsov played 21:22 in the very first game this season, more than any regular season game he played last season. Since this was before he had established any sort of consistency this season, doesn't this suggest that his increased TOI was, at least initially, due more to circumstances beyond Kuznetsov's control? If Backstrom had been healthy this season do you think he would be getting the same opportunities on the PK, for instance? I doubt it.

I also don't buy that his increased TOI dilutes his points/60. There's no data to suggest that this happens with players around the league who are thrust into bigger roles. If Kuznetsov's minutes were reduced to what he received last year, do you think his points/60 would just increase? Seems unlikely.

I think the team is playing better, particularly Ovechkin, and this certainly impacts the z-score numbers above. But I also think this influences how people are perceiving Kuznetsov's game.

You're going down a rabbit hole and racing right past the point. It doesn't have to be 100% causal such as Lavi noting greater regular season play and THEN playing Kuzy more. This was always a "show me" season for him and we also had a preseason and camp, remember? If Kuzy is doing the right things then and he's reached an understanding with Lavi why would Lavi cut his minutes in game 1 of a new season?

If you think Kuzy has been playing the same as before then you're on an island, and that includes everyone who's commented from the team including Kuzy himself.

 

[twabby]

You're going down a rabbit hole and racing right past the point. It doesn't have to be 100% causal such as Lavi noting greater regular season play and THEN playing Kuzy more. This was always a "show me" season for him and we also had a preseason and camp, remember? If Kuzy is doing the right things then and he's reached an understanding with Lavi why would Lavi cut his minutes in game 1 of a new season?

If you think Kuzy has been playing the same as before then you're on an island, and that includes everyone who's commented from the team including Kuzy himself.

I think stylistically he's probably a bit different this year but the output is similar to me.

"All progress has resulted from people taking unpopular opinions."

-Adlai Stevenson