The Advanced Stats Thread Episode VII: An Ode to the Sanity of Silverfish

[silverfish]

See the main issue is that those two were viewed as expendable, and that’s the problem in of itself. How/why are talented players like that being used in lesser roles when they’re obviously outperforming them?

How can you project/predict that they will perform against better players. How can you maximize player efficiency using the system the coach practices? Is that system actually efficient and how can it be improved?

Find players who are destroying their usage, and give them a tougher role. Find players who are getting destroyed by their usage, and give them a softer role. The latter is far more important than the former. It's much easier to hypothesize that a player in tough usage will be better in soft usage than vice versa. Although, some players are definitely just bad, regardless of usage. Worst case scenario, you have a guy destroying easy usage, doesn't do as well in tough usage, you put him back where he was and he goes on destroying. Nothing wrong with being a good third-pairing guy.

Or, you build a model that tries to find players similar to the evaluated player and see how their careers progressed....

 

[Mac n Gs]

Find players who are destroying their usage, and give them a tougher role. Find players who are getting destroyed by their usage, and give them a softer role. The latter is far more important than the former. It's much easier to hypothesize that a player in tough usage will be better in soft usage than vice versa. Although, some players are definitely just bad, regardless of usage. Worst case scenario, you have a guy destroying easy usage, doesn't do as well in tough usage, you put him back where he was and he goes on destroying. Nothing wrong with being a good third-pairing guy.

Or, you build a model that tries to find players similar to the evaluated player and see how their careers progressed....

Some dude eyjee was doing that but he died tragically. RIP in peace aygee

But for real, I wonder what an alternative to that would be. How can we improve scouting and modeling with video-tracked data and shooting stats? I remember seeing somewhere that the best scouting method has been shown to be a combo of both eye test and data for NHL prospects

 

[silverfish]

Some dude eyjee was doing that but he died tragically. RIP in peace aygee

Yeah word on the street is he was trying to clean his window AC and electrocuted himself. Sad!

But for real, I wonder what an alternative to that would be. How can we improve scouting and modeling with video-tracked data and shooting stats? I remember seeing somewhere that the best scouting method has been shown to be a combo of both eye test and data for NHL prospects

Yeah, this: http://statsportsconsulting.com/mai...extMiningScoutingNHLDraftAnalyticsFeb2017.pdf

It all works together.

 

[TomasHertlsRooster]

I’m curious what you guys think of Ryan Merkley. He has a negative team relative GF% on a terrible team, but then there’s this. (These numbers are at 5V5)



Personally, as a Sharks fan, the last 4 drafts (where we know a decent amount about the players selected) tell me that if you’re selecting outside of the top-5, you should always pick the player with the highest potential, even if they have “character issues”. Here is how our last 4 drafts went, and how they would have went if we picked the player with the highest potential that was a faller in the draft:

2012: picked Hertl, would have picked Teravainen
2013: picked Mueller, would have picked Shinkaruk
2014: picked Goldobin, would have picked Ho-Sang
2015: picked Meier, would have picked Barzal.

Obviously, the “would have picked” side is more impressive, which makes me inclined to want to just select the player with the highest ceiling, from this point out.

My question is if any of you guys have seen Ryan Merkley play or know of any other advanced metrics that shed some light on a clearly polarizing player. And whether or not you would take him at #21.

 

[Ola]

JT- Ryan Merkley is very talented, and he has an ability to dictate the tempo of the game that many PMDs from junior hockey might not have. In juniors you can have a lot of success with an ability to just drive the puck up ice. But in the pro game the faster you drive up ice the less time you will get, and against NHL teams you don't have much time to start with. To have success as a PMD in the NHL its more about buying yourself time than taking advantage of Ds on their heels. The jump for many PMDs from the CHL to the NHL have been quite big, there really are no marginals for ever in the pros. You hang onto the puck a split second too long and you will be put on your rear and stripped of the puck, and the reason that the CHL PMDs have fallen behind the European, and many NCAA, competitors is that these kids are premiered a little too much from having a high risk appetite and not enough for being patient with the puck and buying time for themselves. I think Merkley sticks out a little in this perspective because he can do just that really well.

Merkley is also an excellent play-maker who has decent PP skills.

I have two issues with him:

Firstly, how many smaller Ds can teams carry? We are starting to see like 7-8 come up every draft year (this year you have like what Boqvist, Hughes, Merkley, Sandelin, Smith and Lundqvist who all are really good prospects). A bunch of kids I don't know too and these guys sometimes develop late. I would have zero concern about lack of size up front, but undeniably size is still a great advantage when defending in many situations. Without any single doubt, all teams can carry afford to carry 1-2 defenders who aren't big. But you reach a limit, can you have 4 defenders who are between 5'9-11? The more common these guys become, the less valuable they will become. Its simple math. If 5-6 makes the league every year its one per team in 5 years and two per team in 10 years. There are already about 50 Ds in the NHL 6' or smaller playing regularly.

Its not something that should have a major affect, but I wouldn't just throw size as a factor out the window when looking at Ds.

Secondly, the kid is still very immature. My guy feeling say that he is just that, immature, rather than an idiot. Young players being immature often matures, that stubbornness can even become an advantage later on. But its of course a risk.

 

[Mac n Gs]

Understanding Statcast metrics: Deconstructing Hit Probability Understanding Statcast metrics: Deconstructing Hit...

Go get a free trial and read this article about hit probability in baseball. It has a link to a sabermetrics primer too. I’ll post the intro before it dives into the nitty gritty analysis. Hockey needs this type of information that can be brought to the mainstream. I know some people hated it, but the Glopuck could at least help a novice fan follow the game if they aren’t used to following the speed of the game. IIHF has something similar to statcast, so I see no reason this can’t be installed in NHL arenas. If the NHL is smart about their branding/marketing (lol), they could use this as a means to add some intrigue to the more popular players. I hate Pierre, but I’d be interested in hearing him update some of these things in real time: “Wow Edzo, did you see they rocket of a shot by Jared Spurgeon!? Jeff Truitt out of the Kelowna Rockets really coached these defensemen well and that puck flew off Spurgeon’s stick at 120 mph!”

Imagine havint exit velocity off Ovechkin and Laine one-timers? f*** me up, fam.

The times they are a changin’ … and so are the stats. The old process of judging baseball players and teams relied on simple counting stats (runs batted in, hits, wins) and basic rate stats (batting average, fielding percentage). The new era of baseball is defined by hyper-focused metrics based on real-time tracking data provided by Statcast’s array of radar and high-speed cameras installed in every MLB ballpark. Instead of recording only the outcome of a pitch or batted ball, we can now analyze what caused that final outcome.
New things aren’t good by default. Nor are they bad because they’re unfamiliar. Just because we have new data to manipulate, query and analyze, that doesn’t mean the new information is perfect (it’s not), or that it will reveal the secret truths behind the game. Baseball’s objective truth isn’t waiting to be unlocked by one algorithm or projection system. So instead, we use the tools at our disposal. To understand these new metrics, we have to analyze how they are created, what they attempt to evaluate, if they succeed at that intended goal, and how they are presented to the public. Only then can we truly understand them, and learn how to use them responsibly.
In an effort to better educate the masses on these new analytics that have increasingly made their way into mainstream baseball commentary, this is the first in a series of pieces examining Statcast metrics. We begin with Hit Probability, laying out the goals, limitations, misconceptions and calculations that come with one of Statcast’s most derided and misunderstood stats. If you have questions or feedback after reading, please let me know in the comments, where we can continue the discussion.

 

[ManUtdTobbe]

Hey @silverfish , how does Matt Benning look in your tool thingy? I like him a lot but i want backup

 

[silverfish]

Hey @silverfish , how does Matt Benning look in your tool thingy? I like him a lot but i want backup

A lot of nice names here...

Mezsars - 0708
Hamilton - 1516
E. Johnson - 1112
Hickey - 1314
Shattenkirk - 1112
Burns - 0910
Klefbom - 1617
Krug - 1516
Edmunson - 1718
Rielly - 1617

@JoeThorntonsRooster seems very boom or bust to me.

 

[ManUtdTobbe]

A lot of nice names here...

Mezsars - 0708
Hamilton - 1516
E. Johnson - 1112
Hickey - 1314
Shattenkirk - 1112
Burns - 0910
Klefbom - 1617
Krug - 1516
Edmunson - 1718
Rielly - 1617

@JoeThorntonsRooster seems very boom or bust to me.

Ohhh, that's a really nice list... I like him even more now… Gorton should just call Chiarelli every day about Klefbom and Benning until he folds.

 

[Ola]

Understanding Statcast metrics: Deconstructing Hit Probability Understanding Statcast metrics: Deconstructing Hit...

Go get a free trial and read this article about hit probability in baseball. It has a link to a sabermetrics primer too. I’ll post the intro before it dives into the nitty gritty analysis. Hockey needs this type of information that can be brought to the mainstream. I know some people hated it, but the Glopuck could at least help a novice fan follow the game if they aren’t used to following the speed of the game. IIHF has something similar to statcast, so I see no reason this can’t be installed in NHL arenas. If the NHL is smart about their branding/marketing (lol), they could use this as a means to add some intrigue to the more popular players. I hate Pierre, but I’d be interested in hearing him update some of these things in real time: “Wow Edzo, did you see they rocket of a shot by Jared Spurgeon!? Jeff Truitt out of the Kelowna Rockets really coached these defensemen well and that puck flew off Spurgeon’s stick at 120 mph!”

Imagine havint exit velocity off Ovechkin and Laine one-timers? **** me up, fam.

To be honest, I would bet that many NHL teams have this already. I would be surprised if we didn't. Like 2nd tier teams in Swedish Soccer has it, its not even expensive once you can license it.

The only reason for why teams wouldn't have it would be if the programming for soccer took a super effort founded by the rich teams that now has spread out and it has become as common as doing video, but no hockey team has the resources to do the same. But like IIHF got it now, must be coming from some place. The NHL teams have the most resources.

But imagine the potential when you can establish this. You can literary look at everything. There is an extreme amount of details you can look into. Refine your style/system/usage.

 

[Mac n Gs]

To be honest, I would bet that many NHL teams have this already. I would be surprised if we didn't. Like 2nd tier teams in Swedish Soccer has it, its not even expensive once you can license it.

The only reason for why teams wouldn't have it would be if the programming for soccer took a super effort founded by the rich teams that now has spread out and it has become as common as doing video, but no hockey team has the resources to do the same. But like IIHF got it now, must be coming from some place. The NHL teams have the most resources.

But imagine the potential when you can establish this. You can literary look at everything. There is an extreme amount of details you can look into. Refine your style/system/usage.

The NHL is getting ready to implement smart pucks with data-tracking technology

You sure about that? Unless there's some preliminary work being done behind the scenes, this doesn't appear to be implemented league-wide. Maybe some teams pay for it privately?

 

[silverfish]

We give way too much credit to front offices and I doubt that 80% of them aren't doing anything more than logging on to Corsica.

Remember when the Bruins were interviewing GM candidates, and Capgeek had gone down (RIP, Matt Wuest), and some of the people they were interviewing couldn't speak to the Bruins cap situation accurately?

 

[silverfish]

I see that the QoC argument is in full effect again these days. Isn't it already decided that: QoT more impactful than QoC, and it's not that it doesn't matter, but the spread between players is, at a macro level, very small?

 

[Mac n Gs]

I see that the QoC argument is in full effect again these days. Isn't it already decided that: QoT more impactful than QoC, and it's not that it doesn't matter, but the spread between players is, at a macro level, very small?

I wonder how the argument would look if it was analyzed as a WAR/GAR QoC argument. Like what % of your shifts are against a +3WAR player

 

[silverfish]

I wonder how the argument would look if it was analyzed as a WAR/GAR QoC argument. Like what % of your shifts are against a +3WAR player

Would make sense but I think we're years away from a WAR/GAR metric that everyone agrees on

 

[Mac n Gs]

Would make sense but I think we're years away from a WAR/GAR metric that everyone agrees on

Agreed, and I think more people would be receptive if QoC were framed in that manner because every single QoC argument boils down to “QoC really doesn’t matter” and “Oh so you’re saying playing more regularly vs Connor McDavid vs Kevin Hayes doesn’t matter k!”

I just don’t think we have good QoC metrics yet. We don’t have a lot of good metrics yet, so the NHL reallt needs to get their asses in gear and stop being a decade behind the curve.

 

[silverfish]

Agreed, and I think more people would be receptive if QoC were framed in that manner because every single QoC argument boils down to “QoC really doesn’t matter” and “Oh so you’re saying playing more regularly vs Connor McDavid vs Kevin Hayes doesn’t matter k!”

I just don’t think we have good QoC metrics yet. We don’t have a lot of good metrics yet, so the NHL reallt needs to get their asses in gear and stop being a decade behind the curve.

I still think a lot of the things that we think we need will be rendered useless. Player and puck tracking isn't going to tell us a whole lot, and I'm worried about the man hours that are going to be dedicated to deciphering that data to gain 0.5% of a competitive advantage.

Not to mention that every national broadcast is basically "YOU WIN WITH GUYS LIKE CALLAHAN I DON'T CARE ABOUT ANALYTICS".

A sport for dinosaurs.

 

[Mac n Gs]

I still think a lot of the things that we think we need will be rendered useless. Player and puck tracking isn't going to tell us a whole lot, and I'm worried about the man hours that are going to be dedicated to deciphering that data to gain 0.5% of a competitive advantage.

Not to mention that every national broadcast is basically "YOU WIN WITH GUYS LIKE CALLAHAN I DON'T CARE ABOUT ANALYTICS".

A sport for dinosaurs.

I think it’ll help more with model creation and improving xGF metrics. I completely don’t trust someone like Pierre to communicate it efficiently to the public, hence why I mentioned the exit velocity for a slap shot and other cool shit. People will enjoy that more while the other aggregate data can be combed over by the nerds to tell us which players are useful and not. Tracked data will be great for pylons that back up through the boards in an attempt to defend zone entries while measuring how fast the attacking player is getting ready to zoom by them.

 

[Mac n Gs]

What the hell is going on with Manny’s WAR model that Callahan and Vesey have higher WAR than Hedman and McDonagh. Girardi has higher WAR than Skjei. Same applies for his GAR.

CMHockey’s GAR model is so drastically different, so I’m wondering where the disconnect is.

 

[silverfish]

What the hell is going on with Manny’s WAR model that Callahan and Vesey have higher WAR than Hedman and McDonagh. Girardi has higher WAR than Skjei. Same applies for his GAR.

CMHockey’s GAR model is so drastically different, so I’m wondering where the disconnect is.

From what I remember it leans very heavily to xG and offense. Which I think makes sense. Vesey's ixFSh% is ridiculously high so I imagine that influences his oWAR.

But I obviously can't speak to specifics.

 

[Mac n Gs]

From what I remember it leans very heavily to xG and offense. Which I think makes sense. Vesey's ixFSh% is ridiculously high so I imagine that influences his oWAR.

But I obviously can't speak to specifics.

That would make sense, but the results are still a little wonky, IMO. The other model makes a lot more sense when I saw his breakdown and inputs

 

[silverfish]

MAF coming back down to Earth. Who could've seen this coming?

 

[Machinehead]

You know, while I respect what they're doing, I still don't think Washington is that great. Nor is Vegas.

But let's blow it up and go on a 5-10 year plan.

 

[Oscar Lindberg]

what up bois we're being represented in a main thread

 

[Irishguy42]

what up bois we're being represented in a main thread

Oh? I just logged on. What have we done now?