The Advanced Stats Thread Episode VIII:

[Irishguy42]

I think they're borderline too messy at this point and I might ditch the radar chart - but they're basically my trademark so I'll stick with them for now.

I think the differentiation between the lines on the chart could be better. Or, find a way to do less with more, because it's...what....10 lines on one radar?

Edit: I obviously meant do more with less lol

 

[Blue Blooded]

I think the differentiation between the lines on the chart coild be better. Or, find a way to do less with more, because it's...what....10 lines on one radar?

I might lose the SD+, SD- lines for the projections - but they help visualize the general accuracy of that particular projection model. Although they are basically the same for all metrics apart from G60 for defencemen (which is hugely noisy) - so they might go.

 

[silverfish]

Alright, @Blue Blooded, I ran my fancy 'algorithm' on ADA finally. Top comparables:

Tyson Barrie, 25.06
Brady Skjei, 24.2
Yannick Weber, 24.15
John Moore, 23.19
Shayn Gostisbehere, 22.86
Eric Gelinas, 22.5
Shea Weber, 21.18
David Rundblad, 19.24
Keith Yandle, 18.5
Damon Severson, 16.4

All evaluations are done on player seasons +/- 1 year to player current age (so 21-23 year old seasons). A score of 100 is a perfect match (this never happens). A score of 0 represents the players have an 'average' similarity.

I gotta be honest, I don't totally hate that Barrie comparable, but I still don't see that offense from ADA. Now I'm out here letting bias get in the way of my own tool! Gah!

 

[Mac n Gs]

Weird, I was gonna bring up Barrie as a comparable when people were hoping he’d end up like Shattenkirk

 

[silverfish]

Weird, I was gonna bring up Barrie as a comparable when people were hoping he’d end up like Shattenkirk

Oh so you didn't need to spend hours coding metrics to give you a single numeric output in rating player similarity based on given factors to tell you that ADA and Barrie are similar?

SHOWOFF.

 

[Mac n Gs]

Oh so you didn't need to spend hours coding metrics to give you a single numeric output in rating player similarity based on given factors to tell you that ADA and Barrie are similar?

SHOWOFF.

#EyeTestIsElite

No, but I swear that I had their stats side-by-side while I was running something at work, and I was shocked to see the CHL similarities. Just stick the kid with Skjei and don’t look back.

 

[silverfish]

#EyeTestIsElite

No, but I swear that I had their stats side-by-side while I was running something at work, and I was shocked to see the CHL similarities. Just stick the kid with Skjei and don’t look back.

I don't think enough studies have been done on player trajectory. Love what @Blue Blooded is doing with his projection models. And that was the point of me building MILLER in the first place (my player sim model). Looking at establishing similarity scores and then projecting players off of what their similar peers did.

Or maybe they have and I'm just not paying attention. This seems like a most useful thing to have. Like... Neal Pionk is Cody Ceci and ADA is Tyson Barrie, but who do you think the Rangers value more right now? I'm not saying my model is gospel, because far from, but you know what I mean.

 

[Mac n Gs]

I don't think enough studies have been done on player trajectory. Love what @Blue Blooded is doing with his projection models. And that was the point of me building MILLER in the first place (my player sim model). Looking at establishing similarity scores and then projecting players off of what their similar peers did.

Or maybe they have and I'm just not paying attention. This seems like a most useful thing to have. Like... Neal Pionk is Cody Ceci and ADA is Tyson Barrie, but who do you think the Rangers value more right now? I'm not saying my model is gospel, because far from, but you know what I mean.

Oh, it’s totally an are that needs updated results. I still get pissy about that old data studying a player’s prime because I don’t think it’s using enough context to show how the NHL is shifting to a younger league.

As for the comparison models, I think it’s tough just because I think it needs a lot more data than what’s currently out there, starting at the juniors level. If a prospect scoring system can be integrated and utilized, I think it will greatly help projection models moving forward. Once that baseline is established, we can factor in the other measures that are tracked to help build a more complete model.

Maybe I’m expecting too much, but the NHL seriously needs to find a way to improve from a statistics perspective. Minor league baseball shouldn’t have more advanced metrics and tracking than the NHL.

 

[silverfish]

Oh, it’s totally an are that needs updated results. I still get pissy about that old data studying a player’s prime because I don’t think it’s using enough context to show how the NHL is shifting to a younger league.

As for the comparison models, I think it’s tough just because I think it needs a lot more data than what’s currently out there, starting at the juniors level. If a prospect scoring system can be integrated and utilized, I think it will greatly help projection models moving forward. Once that baseline is established, we can factor in the other measures that are tracked to help build a more complete model.

Maybe I’m expecting too much, but the NHL seriously needs to find a way to improve from a statistics perspective. Minor league baseball shouldn’t have more advanced metrics and tracking than the NHL.

At the same time though, I do feel the complaints from some of the crowd in the ADA thread, while I wish they'd go about it better. At what point do you throw out a player's junior numbers when evaluating the player?

 

[Mac n Gs]

At the same time though, I do feel the complaints from some of the crowd in the ADA thread, while I wish they'd go about it better. At what point do you throw out a player's junior numbers when evaluating the player?

It’s tough because it depends player-by-player. I usually try to give them 2 years of professional player to see how they’re scoring translated at the AHL or in limited NHL minutes.

A guy like Colin Miller demolished the AHL in his second year and then went on to post solid stats in limited minutes on Boston. It’s a really inexacy science, but I try to keep it consistentat at roughly a D+4 or D+5 cutoff for dmen. For forwards, I usually do D+3

 

[silverfish]

It’s tough because it depends player-by-player. I usually try to give them 2 years of professional player to see how they’re scoring translated at the AHL or in limited NHL minutes.

A guy like Colin Miller demolished the AHL in his second year and then went on to post solid stats in limited minutes on Boston. It’s a really inexacy science, but I try to keep it consistentat at roughly a D+4 or D+5 cutoff for dmen. For forwards, I usually do D+3

How would you decide on the player-by-player basis, though? I'm not saying you'd do this, but think it opens us up to bias. I like this guy, so I'm going to take his good junior metrics into account longer than this guy I'm ready to write off. Something like that.

I mean, I agree that it's varied, but I'm also not sure myself. With ADA, I think I'm ready to throw out his junior numbers. Not ignore them completely, but not necessarily weigh them heavily. This is a three-year pro already now.

It's a blurry line, and I have no idea where to draw it.

 

[Mac n Gs]

How would you decide on the player-by-player basis, though? I'm not saying you'd do this, but think it opens us up to bias. I like this guy, so I'm going to take his good junior metrics into account longer than this guy I'm ready to write off. Something like that.

I mean, I agree that it's varied, but I'm also not sure myself. With ADA, I think I'm ready to throw out his junior numbers. Not ignore them completely, but not necessarily weigh them heavily. This is a three-year pro already now.

It's a blurry line, and I have no idea where to draw it.

Yeah that’s the huge problem for me when I look at them. It usually ends up being dependent on where they play hockey, and I’ll give a kid out of juniors another year vs a kid in college. I usually see that if you’re not a good college player by your sophomore season, you’re probably not going to be a useful prospect since you’re already on the older end of the spectrum.

This shit is hard man. I’m sure if I’d figured this out by now, y’all would be posting about me on here talking about how I pick reaches in the first round

 

[Mac n Gs]

Totally not overreacting to a 2 and 4 game sample size, but Lias and Chytil posted great shot assist and transition numbers at the end of the year.

 

[silverfish]

Totally not overreacting to a 2 and 4 game sample size, but Lias and Chytil posted great shot assist and transition numbers at the end of the year.

Any numbers on CM or OT?

 

[Mac n Gs]

Any numbers on CM or OT?

Only a one game sample for CM where he generated more shots, minimal shot assists, and was successful on his few zone exit and entry attempts.

Tippett was a volume shooter that got the puck up ice and into the zone effectively (5game SS). Fits his M.O. of basically shooting whenever he deemed fit. He never really was a good puck distributor.

 

[silverfish]

Only a one game sample for CM where he generated more shots, minimal shot assists, and was successful on his few zone exit and entry attempts.

Tippett was a volume shooter that got the puck up ice and into the zone effectively (5game SS). Fits his M.O. of basically shooting whenever he deemed fit. He never really was a good puck distributor.

I f***ing love OT. That dude is gonna score 25 a year and shoot 200 times and will just be a joy.

 

[Mac n Gs]

I ****ing love OT. That dude is gonna score 25 a year and shoot 200 times and will just be a joy.

Yep, wouldn’t shock me at all. He does have his warts though on the defensive end of the puck, so I wouldn’t be shocked if he’s viewed in the same light as Phil Kessel. I highly doubt he’ll be as proficient of a scorer as Kessel though.

 

[silverfish]

Yep, wouldn’t shock me at all. He does have his warts though on the defensive end of the puck, so I wouldn’t be shocked if he’s viewed in the same light as Phil Kessel. I highly doubt he’ll be as proficient of a scorer as Kessel though.

I'd agree with that. His strength is in his volume. I wouldn't call him a pure finisher by any means, and really the only reason he'll be able to pot 25 is because of how much he shoots. Compared to a guy like DeBrincat, who has that volume shooting skill as well as the finishing skill.

Love me them shooters.

 

[silverfish]

Doing some work to clean up the ultra-bad takes in the ADA thread. It's amazing that a large amount of the most vocal posters in there are saying that ADA needs to be a 40 point d-man this year or it's over for him. Does no one realize how many points 40 is for a d-man?

Last few seasons, the number of 40 point d-men, and the number of 40 point d-men who skated < 20 per game

17-18: 30 (3)
16-17: 23 (2)
15-16: 26 (2)
14-15: 27 (1)

It's amazing how scalding these takes can be so far removed from reality. How is a guy who is going to get, likely, 3rd pairing EV minutes and 2nd unit PP minutes, 40 points? It's basically an impossibility. And it's an utter-failure of humanity that we can't get people to move to rate stats. Really, what we should be looking for from ADA is outplaying his usage in terms of shot-attempts, and scoring at the rate of a 2nd pairing d-man. Last year, here were the rates for pairings.

[Data via Corsica.Hockey, 5v5 17-18 minimum 250 minutes]
Pairings are established by using TOI/gp percentiles (ie. top 33% of players are '1st pairing').

1st pairing: 0.85
2nd pairing: 0.77
3rd pairing: 0.68

ADA last year: 0.27

If he operated at a 0.77 p/60 pace, he'd have had 5.7 points (so 6) last year.

Again, really can't believe that rate metrics are still voodoo.

 

[silverfish]

Rochester conference this weekend. Should be getting some cool things released.

 

[Levitate]

Doing some work to clean up the ultra-bad takes in the ADA thread. It's amazing that a large amount of the most vocal posters in there are saying that ADA needs to be a 40 point d-man this year or it's over for him. Does no one realize how many points 40 is for a d-man?
.

I think he needs to show he has the skill to do it, even if he doesn't do it this year. And I say that mostly because I don't really believe NHL coaches are yet cool with the idea of a guy who's main strength is moving the puck up the ice and facilitating other people getting shots and scoring chances if it also doesn't translate into points AND he's not considered strong in his own end.

I can see DeAngelo being a very effective player in the advanced stats arena but struggling to put together a consistent NHL career because coaches aren't yet willing to see the good things he can bring without points to go with them.

 

[silverfish]

I think he needs to show he has the skill to do it, even if he doesn't do it this year. And I say that mostly because I don't really believe NHL coaches are yet cool with the idea of a guy who's main strength is moving the puck up the ice and facilitating other people getting shots and scoring chances if it also doesn't translate into points AND he's not considered strong in his own end.

I can see DeAngelo being a very effective player in the advanced stats arena but struggling to put together a consistent NHL career because coaches aren't yet willing to see the good things he can bring without points to go with them.

I think I agree with this take. He'll continue bouncing around the league and 3 or 4 years from now we'll look at his aggregated stats and wonder why he never got a real shot.

 

[silverfish]

Something we all need to work on. Great point.

Not hockey related, but this presentation from Ron Yurko on developing a WAR model for the NFL was really cool (@stat_ron).

Running list of new people I'm following on Twitter because of today:

@stat_ron (football)
@friscojosh (football)
@PFF_EricEager (football)
@deepfriar (hockey)
@AFPanalytics (hockey)



Another one of those things that is probably easy to fix and extremely effective.

When you get a shoutout during the women's hockey analytics presentation

 

[Mac n Gs]

Is there a replay or summary from the conference available?

 

[silverfish]