The Advanced Stats Thread Episode IX

[Irishguy42]

I'll prob get slammed for this and I have nothing to back it up with apart from anecdotes but WSH is a team who's shot metrics are not commensurate with their time of possession in the offensive zone and their quality of chances aren't properly quantified via the current expected Goal model employed by Corsica.

I really have no concerns about the underlying numbers, their play is sustainable and they have a bunch of elite players, shot attempt%'s be damned.

Yeah, this is why I would love for actual zone times to be tracked.

Puck and jersey tracking as public data, please and thank you.

 

[silverfish]

 

[Machinehead]

I'll prob get slammed for this and I have nothing to back it up with apart from anecdotes but WSH is a team who's shot metrics are not commensurate with their time of possession in the offensive zone and their quality of chances aren't properly quantified via the current expected Goal model employed by Corsica.

I really have no concerns about the underlying numbers, their play is sustainable and they have a bunch of elite players, shot attempt%'s be damned.

You might be right, but I still say who cares about possession if it doesn't lead to shots.

 

[Machinehead]

With Shattenkirk hurt, they went right back to Pionk getting shelled in usage.

Went as well as it always does.

 

[sbjnyc]

You might be right, but I still say who cares about possession if it doesn't lead to shots.

Isn't it a form of shot supression? If you think of shot attempts for and against as a function of TOI then TOI with possession leads to shot attempts for at rate x/60 and TOI without possession leads to shot attempts against at rate y/60. If your puck possession rate is p% then CF% doesn't measure x/(x+y) but rather x/[x-(1-1/p)y]. Does measuring p% add information or is measuring x and y sufficient?

 

[Machinehead]

Isn't it a form of shot supression? If you think of shot attempts for and against as a function of TOI then TOI with possession leads to shot attempts for at rate x/60 and TOI without possession leads to shot attempts against at rate y/60. If your puck possession rate is p% then CF% doesn't measure x/(x+y) but rather x/[x-(1-1/p)y]. Does measuring p% add information or is measuring x and y sufficient?

The Caps are bad at shot suppression.

I never believed that p% tells us anything by itself. It has to do something for you.

 

[sbjnyc]

The Caps are bad at shot suppression.

I never believed that p% tells us anything by itself. It has to do something for you.

If corsi is a proxy for possession shouldn't actual possession matter? If a player maintains the puck longer by, say, not making bad passes right in front of his goalie, I think that'd ultimately improve his CF%.

 

[Machinehead]

If corsi is a proxy for possession shouldn't actual possession matter? If a player maintains the puck longer by, say, not making bad passes right in front of his goalie, I think that'd ultimately improve his CF%.

But it doesn't in this case, that's the thing.

The Caps are not a good corsi team.

I would argue that possession isn't worth much if it isn't leading to measurable success.

 

[Machinehead]

Are we concerned that Zibanejad is a massive GF% drag despite having elite CF% and xGF%?

His usage is absolutely brutal FWIW.

Is that something he's doing? Is it his teammates? Is it simply a consequence of constantly being out against snipers?

 

[Irishguy42]

Are we concerned that Zibanejad is a massive GF% drag despite having elite CF% and xGF%?

His usage is absolutely brutal FWIW.

Is that something he's doing? Is it his teammates? Is it simply a consequence of constantly being out against snipers?

If 2/3 are elite, then is he really the problem?

If he's creating chances and those chances have a high expected goal rate, but his actual goals is low...then eventually life will correct itself and he'll start scoring more goals, which will balance out the goals being scored against him. Assuming he keeps up current play ofc.

I'm going to assume it's the revolving door of wingers he's had lately, plus a dash of poor luck.

High CF%/xGF% and low GF% screams to me: "he's generating a lot of quality, not getting it in, and the opponent is capitalizing on the smaller quantity of chances against."

Doesn't seem weird. Seems like bad luck.

 

[Machinehead]

If 2/3 are elite, then is he really the problem?

If he's creating chances and those chances have a high expected goal rate, but his actual goals is low...then eventually life will correct itself and he'll start scoring more goals, which will balance out the goals being scored against him. Assuming he keeps up current play ofc.

I'm going to assume it's the revolving door of wingers he's had lately, plus a dash of poor luck.

High CF%/xGF% and low GF% screams to me: "he's generating a lot of quality, not getting it in, and the opponent is capitalizing on the smaller quantity of chances against."

Doesn't seem weird. Seems like bad luck.

It only becomes a problem is he sustains that. 2/3 are elite, but at a certain point, the object of the game is to outscore the opponent.

We're not at that point yet.

 

[Irishguy42]

It only becomes a problem is he sustains that. 2/3 are elite, but at a certain point, the object of the game is to outscore the opponent.

We're not at that point yet.

Again, I'd look to teammates here, assuming you're talking about non-rel stats here.

 

[Machinehead]

Again, I'd look to teammates here, assuming you're talking about non-rel stats here.

No that's rel, that's why I brought it up.

 

[Irishguy42]

No that's rel, that's why I brought it up.

Intriguing.

I mean, could still be bad luck. Would have to check the tape and look at the goals against when he's on the ice and the chances he's creating but not getting in.

 

[Ola]

“There’s a bunch of pieces that go into (the decision that a defenseman is ready),” Drury added. “The eye test is one. But we keep our own internal data and notes on every player, Libor included, and we have a lot of people that spend time in Hartford and around the American League watching. We certainly spend a lot of time with all our players down there but he’s one that we keep a close eye on, and Keith and Joe (Mormina, the assistant coach) are career defensemen who played in the American League and know what it’s about. They’ve done a real good job with him and I think each week he’s gotten better and better.”

As Rangers' young forwards make NHL mark, top defenseman...

 

[Ola]

The advanced stats community has to address Washington.

22nd in CF%
27th in xGF%
3rd in GF%

This isn't SSS anymore. They do this year after year. I have always believed that the object of the game is to score goals, and over very large samples, a player is what the GF% says he is, as long as those numbers are consistent. I'm becoming concerned that Erik Karlsson is a consistent goals drag, despite seemingly doing everything right. But for the sake of brevity, let's leave Karlsson alone for now. Derek Stepan is a good example, too. Nobody will accuse him of being Patrice Bergeron by CF%/xGF% but he's a GF% boom for his team year after year.

Can we apply the same logic to an entire team/core? Moreover, the question I've always applied to those players is: what exactly are they doing to influence the GF% a certain way, if the shots and chances are implying the opposite influence?

Shots are shots. There's no two ways about that. Now, is it possible that our existing quality metrics are way off?

Granted, the Caps, unlike last year, are doing ok in K-rating (9th overall). But according to Manny's numbers, their quality metrics are the worst in the league. They succeed, according to Manny's numbers, because they're other-worldly at shooting. I mean, that I buy, because they have the greatest shooter who ever lived. But that doesn't explain why their goals against per 60 is top 5 in the league over the last 3 years. This is not a team with particularly outstanding goaltending.

To be fair, it's only 13th over 17-18 and 18-19, but the quality metrics don't indicate that they should be 13th. They indicate that they should be towards the bottom. And even so, Tampa and Toronto have scored a lot more goals, so it's not like they're just beating everyone to death. This team has a Cup and multiple first-place finishes.

We've missed something here.

My theory is in brief that you attack worse the more you have to worry about what will happen when the team you are facing get the puck. When AO is on the ice you worry a lot...

I am sure of it as a correct theory per se, but its of course hard to try to speculate on the extent of the impact. But like when you face a flat team its so easy to just push harder and harder because you don't have to worry about what will happen if you get a bounce against you. If you instead face a team that often will make you pay every time you make a mistake, its easy to instead lose focus. Isn't there a saying that the Yankees win some games thanks to their stripes? Its also established in soccer, from a betting perspective, that the top clubs in Europe face a lot more upsets in the European Tournaments (Champions League etc) than in their national leagues. Like Manchester City almost never loose against a struggling team in the Premier League, but more of don't get 3 pts against walk overs in inter-continental tournaments.

For various reasons, I think Washington is a team that you just aren't that comfortable to play against. And if you look at how players actually feel when playing Washington away compared to say Florida away -- I think the difference is substantial.

 

[Ola]

“There’s a bunch of pieces that go into (the decision that a defenseman is ready),” Drury added. “The eye test is one. But we keep our own internal data and notes on every player, Libor included, and we have a lot of people that spend time in Hartford and around the American League watching. We certainly spend a lot of time with all our players down there but he’s one that we keep a close eye on, and Keith and Joe (Mormina, the assistant coach) are career defensemen who played in the American League and know what it’s about. They’ve done a real good job with him and I think each week he’s gotten better and better.”

As Rangers' young forwards make NHL mark, top defenseman...

What internal data do we get from the AHL? We have a lot of people following HFD gathering information for our decision making. Good to hear.

 

[Filthy Dangles]

What internal data do we get from the AHL? We have a lot of people following HFD gathering information for our decision making. Good to hear.

It could be literally anything. I'd think it's more scouting notes than a plethora of advanced statistics being logged and recorded. I'd assume things like decision making, did he make the right option or play on a breakout, things like puck races and battles perhaps. That would be my guess.

 

[silverfish]

 

[SnowblindNYR]

So I analyzed the draft. Everyone always says that you can't get elite talent outside the top 5. So I tried to test that. I used the Chi-squared test. I looked at all drafts from 1990-2015 and players that were named to at least one "All-Star Team" and/or the HOF. That's a bit of a better measure than just being an all-star. In case someone else is as ignorant as I am, I didn't know what the difference was, but it appears being named to the All-Star Team is equivalent to an NFL All-Pro. Anyway, the chi-squared test looks at a crosstab and compares the actual to the expected if the crosstab were proportional. Then if the chi-squared provides a p-value of <0.05 then the actual is significantly different from the expected, meaning there's a less than 5% chance that the difference due to random variation.

	AST/HOF	Not AST/HOF	Total			AST/HOF	Not AST/HOF	Total			AST/HOF	Not AST/HOF	Total
Top 5	32	98	130		Top 5	32	98	130		6-10	5	125	130
6-10	5	125	130		Rd 1 NT5	27	575	602		11+ Rd 1	17	455	472
Total	37	223	260		Total	59	673	732		Total	22	580	602
	AST/HOF	Not AST/HOF	Total			AST/HOF	Not AST/HOF	Total			AST/HOF	Not AST/HOF	Total
Top 5	18.5	111.5	130		Top 5	10.5	119.5	130		6-10	4.8	125.2	130
6-10	18.5	111.5	130		6-10	48.5	553.5	602		11+ Rd 1	17.2	454.8	472
Total	37	223	260		Total	59	673	732		Total	22	580	602
P-value	1.64399E-06				P-value	2.07083E-14				P-value	0.895356853

[TBODY] [/TBODY]

Expected is calculated as follows:

For example, in the first chart, for Top 5 players that have made an all-star team or HOF, you can either divide all top 5 picks (130) by all top 10 picks (260) and multiply it by all AST/HOF players selected in the top 10 (37) divided by all players selected in the top 10 (260) and multiply it by all players (260). (130/260)*(37/260)*260. Or more simply, (130*37)/260.

The first test is the top 5 compared to picks 6-10, the next test is top 5 compared to not top 5 picks in the first round, and the last one is 6-10 compared to not top 10 picks. The first two are highly significant, the top 5 is disproportionately strong in these types of players compared to the bottom half of the top 10 and compared to the entire first round that's not the top 5. However, the bottom half of the top 5 is not any better in producing these type of players than first round picks that are not in the top 10. In fact, even eye-balling it, you'll see an uncanny proportionality. Each expected value rounds to the actual. And the p-value is extremely high.

 

[silverfish]

 

[Machinehead]

 

[TomasHertlsRooster]

The advanced stats community has to address Washington.

22nd in CF%
27th in xGF%
3rd in GF%

This isn't SSS anymore. They do this year after year. I have always believed that the object of the game is to score goals, and over very large samples, a player is what the GF% says he is, as long as those numbers are consistent. I'm becoming concerned that Erik Karlsson is a consistent goals drag, despite seemingly doing everything right. But for the sake of brevity, let's leave Karlsson alone for now. Derek Stepan is a good example, too. Nobody will accuse him of being Patrice Bergeron by CF%/xGF% but he's a GF% boom for his team year after year.

Can we apply the same logic to an entire team/core? Moreover, the question I've always applied to those players is: what exactly are they doing to influence the GF% a certain way, if the shots and chances are implying the opposite influence?

Shots are shots. There's no two ways about that. Now, is it possible that our existing quality metrics are way off?

Granted, the Caps, unlike last year, are doing ok in K-rating (9th overall). But according to Manny's numbers, their quality metrics are the worst in the league. They succeed, according to Manny's numbers, because they're other-worldly at shooting. I mean, that I buy, because they have the greatest shooter who ever lived. But that doesn't explain why their goals against per 60 is top 5 in the league over the last 3 years. This is not a team with particularly outstanding goaltending.

To be fair, it's only 13th over 17-18 and 18-19, but the quality metrics don't indicate that they should be 13th. They indicate that they should be towards the bottom. And even so, Tampa and Toronto have scored a lot more goals, so it's not like they're just beating everyone to death. This team has a Cup and multiple first-place finishes.

We've missed something here.

I have been saying this for a while now. From 2015-2016 to 2017-2018, their 5V5 GF% is 56.57%. The next best team is Tampa Bay at 53.78%.

And it’s not just Ovechkin. Even if you remove every goal they got from Ovechkin, the league’s leading 5V5 goal scorer over that time frame, their GF% drops to 52.99%. That would be 5th in the NHL! For reference, if you remove Nikita Kucherov’s goals from 2nd place Tampa, their GF% drops to 50.42%.

I recall that while I was lurking this thread during the playoffs, somebody re-posted an xGF% model that favored them winning against Tampa? I’d love to see that.

 

[Mac n Gs]

I have been saying this for a while now. From 2015-2016 to 2017-2018, their 5V5 GF% is 56.57%. The next best team is Tampa Bay at 53.78%.

And it’s not just Ovechkin. Even if you remove every goal they got from Ovechkin, the league’s leading 5V5 goal scorer over that time frame, their GF% drops to 52.99%. That would be 5th in the NHL! For reference, if you remove Nikita Kucherov’s goals from 2nd place Tampa, their GF% drops to 50.42%.

I recall that while I was lurking this thread during the playoffs, somebody re-posted an xGF% model that favored them winning against Tampa? I’d love to see that.

IIRC, it was Micah McCurdy

 

[SnowblindNYR]

Glad my post got completely ignored.