The Heavy Lifter Index
by SO_RyanP on Jan 5, 2011 9:08 PM EST
Email
Print
As a way to measure which players in the NHL are carrying the heaviest loads and outperforming their competition, I created the Heavy Lifter Index on Hockey Prospectus.
The idea was largely inspired by some of Gabe's posts here at BTN and also by the work of Derek Zona over at Copper n' Blue. My hope was to find some type of index that could show the relative rarity of a players situational stats and performance stats combined.
Part 1 - Explains the methodology
Part 2 - Refines my method and delves into HLI for forwards for the last 3 seasons
If you're interested in the topic, please give it a read. Players like Mikko Koivu, Patrice Bergeron, Andrew Ladd and Brandon Dubinsky all show quite well in individual seasons while others like Martin Hanzal, Mike Richards and Alex Burrows showed very well across the 3 seasons total. I'm sure you won't be surprised by the top player on the list.
Given the great statistical minds who read this blog, I'd love to hear your feedback in the comments section here or at HP. Any insight into how to improve HLI or use it for further analysis would be great.
26 comments
|
0 recs |
Do you like this story?
Comments
Hey Ryan, nice article. Here are a couple of thoughts…
1) Just to clarify, if a guy has a 40% zone start which is 1 standard deviation below league average, does that count as a +1 or a -1? (I would think +1 based on your description in the article)
2) Why make a cut off for Corsi Rel QoC at all? If a guy has a low Corsi Rel QoC, then he’ll be below the mean and penalized in your formula. For example, if you have a guy who faces the top line and draws even with them per 60, and then you have another guy who faces a line half as good (so he doesn’t make the Corsi Rel cut) but outscores them by two goals per 60, the second player is just as valuable as the first player (assuming a linear relationship). I don’t think taking on tough competition qualifies one as a heavy lifter without more; getting roasted by the better players isn’t something any GM wants.
3) I know it’s been discussed previously on this blog, but I would suggest finding a way to adjust for the quality of teammates.
Hockey Blogger at Pensburgh.com
1) Exactly. It would count as a +1 because that’s a much tougher situation to face.
2) The reason being that I designed the system to measure forwards who faced the hardest competition. I could use the same system to measure all forwards, but there would definitely be some players who’s corsi and scoring differential positives outweighed their situational negatives. They are still very good players, but just aren’t Heavy Lifters.
3) I would like a way to include QoT but from what I read (and agree with) is that since the sampling of players that qualify as teammates isn’t nearly the same as those that qualify as competition, the sample size isn’t as valid. Also, the teammates’ results are influenced greatly by the player being observed. QoT metrics are not nearly as reliable as QoC metrics at this point. That said, once there are better measures, I would definitely include them.
Ryan
Matchsticks & Gasoline, Sports Opinionated, Pink Shirt Wise Guys: Italian Soccer Podcast & occasionally even Hockey Prospectus. Apparently I have commitment problems.
The reason being that I designed the system to measure forwards who faced the hardest competition.
We already can find that out by looking at QoC metrics. Obviously what we really want to know (and what I’m sure you had in mind when making this) was who can face tough competition and still come out positive. I think we can find that out without excluding a large portion of the players in the NHL. Like I said above, a player who gets pummeled on the ice does not deserve to be on this list anymore than a guy who plays weaker lines but can outscore them by 2 or more over 60 minutes. I doubt the top 15 or 20 guys would change much if you took away the arbitrary cutoff.
QoT metrics are not nearly as reliable as QoC metrics at this point.
Where’s the articles on this?
Hockey Blogger at Pensburgh.com
I think Heavy Lifter tries to cover guys like Vanek and Kovalchuk, and Ovechkin this season, who don’t show up as terribly tough competition by Corsi Rel QoC but actually are, at least as a defensive assignment.
As for the second part, it’s more or less intuitive, I’d think. Last year the top Capitals have QoTs in the neighborhood of 5, but this year it’s half that. Much smaller sample of players to choose from, so you’re basically averaging a dozen players for a guy like Ovechkin. There’s no doubt that teammates influence each other heavily when they play together a lot.
My blog and Twitter, featuring coverage of the most frustrating team in the NHL
If you don't know how to use Timeonice, read this.
Behindthenet quick link to QoC/QoT/Corsi/PDO/Zonestarts
"Numbers don't lie, they just don't agree with you"--George E. Ays
by red army line on Jan 10, 2011 9:44 AM EST up reply actions
I’m not sure I understand your response to the first point. Mind elaborating?
And I get what you’re saying for the second point, but it’s dangerous to fawn over Z or Dats when they’re playing with quite possibly the best teammates any center has had the last 5-10 years.
Hockey Blogger at Pensburgh.com
WOWY can help a bit with this kinda thing.
by Passive Voice on Jan 10, 2011 12:01 PM EST up reply actions
Soo…they’re not awesome because they have their awesome linemate to lean on?
"Essentially, all models are wrong, but some are useful" George E.P. Box
by Knee high to a duck on Jan 10, 2011 12:32 PM EST up reply actions
I wonder when the haterz start attributing Crosby’s success to Letang in Hart discussions, and Letang’s to Crosby in Norris discussions (a la Ovechkin-Green-Backstrom).
My blog and Twitter, featuring coverage of the most frustrating team in the NHL
If you don't know how to use Timeonice, read this.
Behindthenet quick link to QoC/QoT/Corsi/PDO/Zonestarts
"Numbers don't lie, they just don't agree with you"--George E. Ays
by red army line on Jan 10, 2011 12:41 PM EST up reply actions
I think it makes Crosby’s numbers less impressive, since both of them are feeding off of one another and benefiting. But Crosby’s wingers are just so terrible that it’s tough to take seriously anyone claiming that his accomplishments should be diminished because of the quality of his teammates.
Hockey Blogger at Pensburgh.com
We both know Z and Dats would have significantly less impressive numbers if they were separated. That much is easy. On top of that, who else gets to skate with players like Franzen, Rafalski, and Lidstrom on a regular basis? Z and Dats both benefit quite a bit from having Hall of Fame talent (at least on the blue line) constantly on the ice with them. I don’t think either of them are not elite because of their teammates, but I think this facet of the argument is too often overlooked.
Hockey Blogger at Pensburgh.com
First point: Kovalchuk is somewhat pedestrian, or even below average, in Corsi Rel. Thomas Vanek doesn’t show up well. This season Ovechkin is in low single digits. That won’t show up as good competition, but you know each team’s top checkers and shutdown duo go up against them every night, no matter what Corsi says. Malkin I believe is in a similar boat. For Kovy at least, his true talent is in shooting accuracy, not volume, so it makes sense to incorporate that into any QoC metric. Corsi Rel QoC doesn’t. I’m sure there are similar players, or players perhaps with good Corsis who can’t score at all (like Scott Gomez).
Just by the way, Dobber Hockey is giving me ~40% of Z’s shifts with Datsyuk this season, and somewhere in the neighborhood of 25-30% last season.
Z (#40) with Dats (#13; Lidstrom is #5). Looks like Z is doing pretty well without Datsyuk. Don’t know about QoC, but here are Z and Dats zone starts (I think I messed up somewhere in the script, though).
Maybe this is the most damning. Lidstrom + Zetterberg + Datsyuk is basically the same as Lidstrom + Zetterberg with no Datsyuk, and I’d bet the quality of competition stays the same. And Z is solid without 5, if I’m reading this right (though I guess QoC would be easier).
Wow, Zetterberg is even better than I thought. Freak.
My blog and Twitter, featuring coverage of the most frustrating team in the NHL
If you don't know how to use Timeonice, read this.
Behindthenet quick link to QoC/QoT/Corsi/PDO/Zonestarts
"Numbers don't lie, they just don't agree with you"--George E. Ays
by red army line on Jan 10, 2011 12:32 PM EST up reply actions
Thanks for that. I think the unfortunate thing about all of this is sample size issues. Z and Dats are both very very good, but we would need multiple seasons where Z and Dats played separately all the time without HoF help at the blue line to really see where they’re at.
But after Crosby, those two are the next best centers (maybe a little homerism in there).
Hockey Blogger at Pensburgh.com
For the first point – I can absolutely include all forwards, but there has to be a filtering system to ensure we stick to the criteria of a Heavy Lifter. That said, the index could definitely be used to measure all forwards.
I’m not necessarily trying to measure who faced the hardest competition. I mis-stated that. I’m trying to measure who was able to perform the best against top competition.
As for the QoT vs QoC thing – I remember reading a post by Gabe that QoT is not as statistically reliable or sound as QoC because it doesn’t inherently use the same sample size and is greatly influenced by the player you’re measuring. For some reason I can’t find the article.
Ryan
Matchsticks & Gasoline, Sports Opinionated, & Hockey Prospectus. My twitter handle is @sprtopinionated
Essentially, this is an adjusted Corsi rating and it looks at scoring differential while on the ice and penalty differential?
Is there any weighting done to this? For example, is a better scoring differential weighted more heavily than a Corsi score? If not, why? Also, would it be useful to include a Pts/60 metric to see who he is also driving the offense even in these tough situations?
Do you have spreadsheet available with the data? How many forwards qualified?
Just eyeballing the data from this year, it would appear that Kesler, Zetterberg and Datstyuk (prior to his injury) are each having great seasons again. Their Corsi and outscoring rates are off the charts considering their zone start %. Kesler’s quality of comp data suggests that he hasn’t been facing the toughs this year, but I’m not sure how accurate that is. I know Zetterberg and Datstuk get the toughs on a nightly basis. The Wings and Canucks appear to have opposite philosophies and perhaps that is dictated by the players on their respective rosters. But, the Wings give their best players the toughest assignments on the team (in terms of O zone % and quality of comp) and shelter their depth players. The Canucks, on the other hand, shelter the Sedins and start them in the O zone as much as possible and feed their other lines the majority of the defensive zone faceoffs. It could make a potential playoff series between the two teams pretty interesting.
It’s not an adjusted Corsi per se. I don’t actually adjust the Corsi rate, I just use standard deviations from the league mean to explain either the rarity of the player’s performance or the rarity of the situations they’re put in.
I kept the weighting equal, but deep down I believe it should be. I just didn’t want to arbitrarily assign a weighting unless I had some statistically sound reason to do so. If anyone’s got a way to show a rational for weighting the 6 factors, I’d love to see it.
I have all the data but not currently with me, I’ll check the total number of forwards that qualififed tonight.
I plan to run HLI for this year’s season pretty soon so we can see who stands out at the halfway point of the season.
Vancouver vs Detroit would be a phenomenal series to watch. I suspect that Kesler’s getting easier icetime due to the addition of Malhotra, but I’d have to check.
Thanks for the comment.
Ryan
Matchsticks & Gasoline, Sports Opinionated, Pink Shirt Wise Guys: Italian Soccer Podcast & occasionally even Hockey Prospectus. Apparently I have commitment problems.
It’s difficult to talk about all the hard work Z and Dats have done when you realize their QTEAM is 29th and 26th respectively out of the 371 forwards who have played 20 games so far. Their Corsi QoT is also 11th and 6th respectively out of the 371 players who have played 20 or more games.
Hockey Blogger at Pensburgh.com
They influence their quality of teammate rating, though. Z and Pavel play together a lot and they carry around Holmstrom and get him a high Corsi score, as well. B/c that line plays together a lot and has a high Corsi score together, their quality of teammate score will reflect this.
Right, but the fact that Z and Dats get to play together is a big knock against the story that they’re playing incredibly tough minutes. Each of them would not look as good if they didn’t have the other to lean on.
Hockey Blogger at Pensburgh.com
Z and Dats were playing separately last season, I’m fairly sure.
My blog and Twitter, featuring coverage of the most frustrating team in the NHL
If you don't know how to use Timeonice, read this.
Behindthenet quick link to QoC/QoT/Corsi/PDO/Zonestarts
"Numbers don't lie, they just don't agree with you"--George E. Ays
by red army line on Jan 10, 2011 9:44 AM EST up reply actions
I know they’re playing together a lot this season though, and I think Babcock has worked them together more often at evens than say Crosby and Malkin.
Hockey Blogger at Pensburgh.com
But, they were not on each others line consistently last season or in 08-09, and they still had great results despite each taking turns taking on tough competition.
QoT isn’t just their wings, it’s also the HoF help they get at the blue line. Both of these guys are very, very good, but I wish we lived in a world where we could get multiple seasons of Z and Dats where they don’t play with each other and don’t get big help from the back end.
Hockey Blogger at Pensburgh.com
It is hard to separate the individual from the team. If using that argument, then one would just say that the Wings top players as a whole (Dats, Z, Lidstrom, etc) do very well against top competition year after year, but attempting to figure out individual contributions is difficult.
I do think it is somewhat interesting, though, that Lidstrom and Stuart take on the tough competition and get a lot of defensive zone draws and play with Z and Dats a lot b/c of that, but their Corsi’s are very different. Lidstrom and Stuart rank near the bottom of the team in Corsi and Z and Dats rank at the top of the team. It would seem that they together are driving the play.
I’d love to see the full rankings to see where Armstrong slotted in every year in order to end up in the top 20 overall as well as mid-season rankings.
Cheers if you have time.
Pension Plan Puppets: A Toronto Maple Leafs blog and a group therapy session.
Like reading thoughts confined to 140 characters? I'm on Twitter too.
Cool work, but I had a question. FTA:
For each player, the standard deviations for all six metrics were summed to create the Heavy Lifter Index. An individual player’s HLI Sit is the sum of the standard deviations for Corsi Rel QoC, QualComp and Zone Starts. A player’s HLI Res is the sum of the standard deviations for Scoring Differential, Corsi and Penalty Differential. Summation of the standard deviations was used to reward players for the degree to which they differed from league norm, rather than just whether they cleared an arbitrary hurdle. We want to know if he cleared that hurdle by two inches or two feet.
Does that not mean that each hurdle was equally weighted? Is there evidence that offensively being 2 standard deviations better on Corsi is as positive for your team as being two deviations better in Penalty Differential?
http://hockeyzen.com
For some reason I’m having trouble with the nested comments today.
PPP – I posted the Colby Armstrong results on your podcast episode 13 comment string
mindmasher – Yes, each hurdle is equally weighted. No, there is no evidence (that I know of) to say that being 2 standard deviations better at Corsi is the same as a similar spread in Penalty Differential. I would love to do a version 2.0 in the future with a weighting factor – but need more info to do that. For the time being I was quite happy that using std dev’s helped me compare inherently different measures but there is still plenty of work to do.
Ryan
Matchsticks & Gasoline, Sports Opinionated, & Hockey Prospectus. My twitter handle is @sprtopinionated
Comments For This Post Are Closed
Secondary Sidebar
User Tools
Related FanPosts on Arctic Ice Hockey
FanPosts
Community blog posts and discussion.
