HFBoards

Go Back   HFBoards > General Hockey Discussion > By The Numbers
Mobile Hockey's Future Become a Sponsor Site Rules Support Forum vBookie Page 2
By The Numbers Hockey Analytics... the Final Frontier. Explore strange new worlds, to seek out new algorithms, to boldly go where no one has gone before.

A little perspective on the current star defenseman

Reply
 
Thread Tools
Old
08-11-2012, 05:35 PM
  #1
Sureves
Registered User
 
Sureves's Avatar
 
Join Date: Sep 2008
Location: Ottawa
Posts: 7,131
vCash: 500
A little perspective on the current star defenseman

Most of us know that there is a fairly consistent consensus as to who are the best defenseman in the NHL, but there are large varieties of opinions when it comes to ranking those stud defenseman.

So what I wanted to do was to look at the top 10 defenseman as voted by HFboards (http://hfboards.hockeysfuture.com/sh....php?t=1243801) and to look at them a little closer in terms of what they contribute to their respective teams in terms of winning games.

I excluded Lidstrom because he's retired, and Yandle because he was so low in Corsi Rel QoC and thus wasn't really easily comparable to the other defenseman in my opinion. I replaced him with OEL for personal interest.

Another thing to note with regards to QoC as I am quite sure this will come up in the posts in this thread, is that QoC plays almost no role in terms of a defenseman's advanced statistics over an 82 game period because they all face top competition, and even if one defenseman faces top competition more than the other, the difference in advanced statistics (ie. goals against, goals for, etc.) is so negligible that it isn't worth analysing (You can read about that here: http://nhlnumbers.com/2012/7/23/the-...of-competition).

As such, because these are all first pairing defenseman who face top competition on a regular basis, we can compare apples to apples here and no one should take issue with that and claim the "tougher competition" card. Please read the article if you are going to make such a post before posting to keep this thread clean.

So the first thing I looked at was overall +/- per 60 minutes of hockey. Naturally this indicates which defenseman was "best" when on the ice though it does not consider the strength of his teammates, or the strength of their team as a whole.



This picture suggests Kris Letang was outstanding when on the ice and is by far the best of the defenseman. Chara is second which should come as no surprise as both Pittsburgh and Boston were great teams finishing #4 and #7 in league standings respectively.

Next is Erik Karlsson, which is quite surprising because his team was #16 in league standings, the lowest standing of all the players' teams in this study.

Of particular note is that Alex Pietrangelo finished 6 of 9 in this metric despite being on the #3 league wide St. Louis Blues. This is potentially attributable to the fact that St. Louis is one of the most balanced teams in the NHL and has a scoring-by-committee type of play and thus are boosted in the standings not from one dominant line, but by multiple strong lines (which would explain why his +/- per 60 may not be as high due to potentially not that strong of teammates relative to some of the other guys in this study).

Next I looked at +/- per 60 relative to their teams. This essentially showed how much better the team did when the defenseman in question was on the ice in terms of +/-. Effectively this illustrates how much this player "boosted" their team over their standard play when he was on the ice. It should be noted that this metric is biased towards players on bad teams because if that team has only one good line (but it is a great line) and that defenseman plays with that line, that will obviously make them look better than they are:



This metric suggests again that Kris Letang is the most dominant defenseman - especially impressive considering his dominance in the previous metric. However, it should be noted that Kris Letang did not play nearly a full year this year, and it's possible that the games he played the team played better than they did on average (possibly as a result of the fact that he was playing in the first place?). Further analysis would be needed to properly account for this, but I will instead say that this is interesting for another study, but not in itself a clear indication that Letang is in fact as dominant as this indicates (though it potentially is).

In 2nd place is Erik Karlsson. As mentioned this is not overly surprising since Ottawa is very much a one-line team where lines #3,4 and to a lesser extent #2, are not nearly as good as line one where Erik Karlsson plays mostly.

Weber is 3rd, which shouldn't be much of a surprise given the fact that he is one of the best defenseman in the game and is also paired with an excellent partner in Ryan Suter who places 5th.

Next is Chara, again no surprise since he is one of the best defenseman in the game.

Of particular interest is that Pietrangelo is #7 of 8 in this metric, which can likely be partially attributed to the fact that St. Louis is such a good team that is widely a scoring by committee affair as discussed earlier.

Lastly, and what I think is the best and is my personal contribution, is a new metric I made which looks at the defenseman's most frequent linemates (top 10) excluding their main defensive partner (so 9). That is, this doesn't merely look at the Team +/-, but rather looks as the players +/- with which the defenseman in question plays with the most.

This effectively allows us to gauge exactly how much better the team performs when he is on the ice relative to his most frequent teammates. This seeks to isolate how good each defenseman really is and is in my opinion the best of the three metrics posted in this thread. I consider this to be the best representation of how good a defenseman is.

It is a weighted average based on % of ice time and the corresponding linemate's +/- ON.

The results are as follows:



Again Letang wins this metric, but again, it should be noted that the +/- ON of his teammates used in the equation is for 82 games, and he didn't play all those games so it's kind of comparing apples to oranges. You'd need to do more work to paint a clearer picture.

Karlsson is second and given my avatar, of course I'm going to talk about this. What this means is that Ottawa performed relatively better than every other team in the study when Karlsson was on the ice after controlling for the fact that he gets to play with Spezza, Michalek, and Greening so much.

In my opinion, this is a pretty huge conclusion given the fact that he won the Norris this year even though many don't think he should have and thought it should have gone to...

Shea Weber. The guy who just so happens to be pretty much neck-and-neck with Karlsson in this metric (in fact, it's basically a tie, I just posted Karlsson first to rattle some peoples' cages ). Again, I believe this metric is a very good measure of how good a defenseman is at causing their team to win games.

Next is Suter, again, not surprising since he's paired with Weber almost all the time so it's hard to conclude from strictly statistics which one is better but it's generally considered to be Weber. Again, it should be noted that Weber probably is being helped a fair bit by Suter in all 3 of the metrics whereas Karlsson is working only with Kuba.

Then of course is our next Norris nominee Mr. Chara who is of course an excellent defenseman and widely considered to be one of the best in the league.

Pretty interesting results in my opinion, feel free to discuss the three and speak your mind as to which (if any) you like.

Sureves is offline   Reply With Quote
Old
08-11-2012, 06:02 PM
  #2
Batali Crocs*
 
Join Date: May 2011
Country: Canada
Posts: 1,163
vCash: 500
Chara's the best well rounded Dman in the league.

Batali Crocs* is offline   Reply With Quote
Old
08-15-2012, 11:21 AM
  #3
Czech Your Math
Registered User
 
Czech Your Math's Avatar
 
Join Date: Jan 2006
Location: bohemia
Country: Czech_ Republic
Posts: 3,665
vCash: 500
Interesting. Do you have any data for previous seasons? I'd really like to see comparisons between players who finished near the top of Norris/AS voting and/or players like Lidstrom, Chara and Pronger.

Czech Your Math is offline   Reply With Quote
Old
08-15-2012, 12:08 PM
  #4
wgknestrick
Registered User
 
Join Date: Aug 2012
Posts: 2,146
vCash: 500
As anyone will tell you, results without context have little value. Qual Comps need to be added to see if Letang posted these results against top competition or not. Being a PIT fan and watching just about every game for the last 5 years, Letang usually draws the 2nd lines and this is verified by him having the 3rd highest Qual Comp on his team.

Zone starts also give great context. Posting great +- is easy when you start 50% of the time in the O zone, like Letang does.

I also suggest looking at +- OFF ICE/60min for the players. This gives you the WYWO perspective.

SV% ON ice while having a high QualComp will generally show you who the studs are (on a team). Just note that you can ONLY compare this to their teammates.

He is an elite Dman (or at least has been for the last 1.5 seasons), but is not on the scale of Chara, Webber, or probably Karlson IMO. I think Karlsson and Letang will probably be the best 2 offensive D men in the league for a while with Green falling off the map. Letang will soon be a 7mil D man or more on his next contract.


Last edited by wgknestrick: 08-15-2012 at 12:18 PM.
wgknestrick is offline   Reply With Quote
Old
08-15-2012, 05:54 PM
  #5
Verviticus
Registered User
 
Join Date: Jul 2010
Posts: 4,989
vCash: 500
qualcomp isn't going to give context. read the article he posted re: QoC - for the most part, his average QoC over the course of a year is going to make a difference by at most a couple %

zone starts would be really useful and important. also note that letang's on ice sh% was almost 12%! this year, and his last two years have been 8.4% and 8.46%. That's a pretty significant amount of his production that's probably riding the percentages

since it's +/-, ie: something entirely reliant on sh% and sv%, you could probably do a really quick adjustment for PDO, see what that gets you, and then try to adjust for zone starts as well. the results would likely be very different and probably more accurate, and I'll bet you'll get a list closer to people's impressions rather than one that puts Letang at the top

edit: I'll add that ideally, you wouldn't just assume a PDO of 1000. you could look at fleury's evsv% historically, letangs average sh% historically and come up with an aggregate that you could then use to adjust the results you have and that would probably be even more accurate

Verviticus is offline   Reply With Quote
Old
08-16-2012, 02:50 PM
  #6
Sureves
Registered User
 
Sureves's Avatar
 
Join Date: Sep 2008
Location: Ottawa
Posts: 7,131
vCash: 500
Quote:
Originally Posted by wgknestrick View Post
As anyone will tell you, results without context have little value. Qual Comps need to be added to see if Letang posted these results against top competition or not. Being a PIT fan and watching just about every game for the last 5 years, Letang usually draws the 2nd lines and this is verified by him having the 3rd highest Qual Comp on his team.

Zone starts also give great context. Posting great +- is easy when you start 50% of the time in the O zone, like Letang does.

I also suggest looking at +- OFF ICE/60min for the players. This gives you the WYWO perspective.

SV% ON ice while having a high QualComp will generally show you who the studs are (on a team). Just note that you can ONLY compare this to their teammates.

He is an elite Dman (or at least has been for the last 1.5 seasons), but is not on the scale of Chara, Webber, or probably Karlson IMO. I think Karlsson and Letang will probably be the best 2 offensive D men in the league for a while with Green falling off the map. Letang will soon be a 7mil D man or more on his next contract.
1. As per Qualcomp article, that will factor very little into the analysis. Probably not worth bothering with since it'll be an immaterial difference.

2. Zone starts are important. True, it would be useful to include them in the analysis no doubt. You'd have to use relative zone starts though (relative to the team you play on, for example, Ottawa is in the offensive zone more than 50% of the time, while St. Louis let's say is in the offensive zone less than 50% of the time, Player A on Ottawa at 50% offensive zone starts is actually being more defensively used than Player A on St. Louis at 50% offensive zone starts.

3. Already captured that in +/- relative (2nd chart)

4. Save percentage ON: I'm not really sure there's a lot of value added to including that in the analysis, but to each his own.

The main issue with Letang is that he didn't necessarily play enough games for Qualcomp not to matter, and it's also comparing apples to oranges in using team-relative stats since the stats he has are from only a sample of the team's total 82 games.

Sureves is offline   Reply With Quote
Old
08-16-2012, 02:53 PM
  #7
Sureves
Registered User
 
Sureves's Avatar
 
Join Date: Sep 2008
Location: Ottawa
Posts: 7,131
vCash: 500
Quote:
Originally Posted by Czech Your Math View Post
Interesting. Do you have any data for previous seasons? I'd really like to see comparisons between players who finished near the top of Norris/AS voting and/or players like Lidstrom, Chara and Pronger.
I'm thinking about doing that next while including a new column for relative offensive zone starts.

Sureves is offline   Reply With Quote
Old
08-16-2012, 05:44 PM
  #8
Verviticus
Registered User
 
Join Date: Jul 2010
Posts: 4,989
vCash: 500
Quote:
Originally Posted by Sureves View Post
2. Zone starts are important. True, it would be useful to include them in the analysis no doubt. You'd have to use relative zone starts though (relative to the team you play on, for example, Ottawa is in the offensive zone more than 50% of the time, while St. Louis let's say is in the offensive zone less than 50% of the time, Player A on Ottawa at 50% offensive zone starts is actually being more defensively used than Player A on St. Louis at 50% offensive zone starts.
i dont think this is true

Quote:
4. Save percentage ON: I'm not really sure there's a lot of value added to including that in the analysis, but to each his own.
it's essentially required. so is sh%, and if you ignore both, the results are very skewed.

Verviticus is offline   Reply With Quote
Old
08-16-2012, 06:28 PM
  #9
Talks to Goalposts
Registered User
 
Join Date: Apr 2011
Location: Edmonton
Country: Canada
Posts: 3,637
vCash: 500
The biggest factor for Letang's huge plus minus was his 11.93 shooting percentage which seems to be due to spending so much time with Malkin and Crosby and getting somewhat less bad goaltending from Fleury than the other Penguins top 4 defenders with .913. I doubt he's the guy driving the bus here rather than being PIT's best defender who spends most of his time with two elite forwards.

If you did this kind of study last year you might conclude Ehrhoff was awesome when really he was just a caddy for the Sedin line. I think this may be a similar situation.

Overall I think you made the mistake of thinking these guys all play similar roles. Most of the guys on the list play general use 1st pairing roles while Letang plays support to Pittsburgh's scoring forwards while Staal, Orpik and Michalek were the primary defensive guys.

Talks to Goalposts is offline   Reply With Quote
Old
08-16-2012, 06:58 PM
  #10
Sureves
Registered User
 
Sureves's Avatar
 
Join Date: Sep 2008
Location: Ottawa
Posts: 7,131
vCash: 500
Quote:
Originally Posted by Verviticus View Post
i dont think this is true



it's essentially required. so is sh%, and if you ignore both, the results are very skewed.
1. You can think it's not true if you want, but it is

2. Using save percentage ON (relative to team) is not a good idea because it's heavily driven by noise in the short run and will regress towards the mean in the long run. Fancy way to say it's mostly luck. Same with shooting percentage. The study done is performed to illustrate who had the better year - regardless of reasons (luck perhaps) - in terms of net positive production, not to determine who is the better overall player in the long-run.

As such, you don't need to account for sv percent on, or shooting percent on.

Sureves is offline   Reply With Quote
Old
08-17-2012, 01:23 AM
  #11
Verviticus
Registered User
 
Join Date: Jul 2010
Posts: 4,989
vCash: 500
Quote:
Originally Posted by Sureves View Post
1. You can think it's not true if you want, but it is
zone starts aren't relative. can you please explain this more thoroughly?

Quote:
2. Using save percentage ON (relative to team) is not a good idea because it's heavily driven by noise in the short run and will regress towards the mean in the long run. Fancy way to say it's mostly luck. Same with shooting percentage. The study done is performed to illustrate who had the better year - regardless of reasons (luck perhaps) - in terms of net positive production, not to determine who is the better overall player in the long-run.

As such, you don't need to account for sv percent on, or shooting percent on.
yeah ok, if this isnt meant to be predicative - but then what's the point? is it a study for its own sake? I guess, what conclusions can we extract from it, and what things can we do with those conclusions?

Verviticus is offline   Reply With Quote
Old
08-17-2012, 10:49 AM
  #12
wgknestrick
Registered User
 
Join Date: Aug 2012
Posts: 2,146
vCash: 500
Quote:
Originally Posted by Verviticus View Post
zone starts aren't relative. can you please explain this more thoroughly?



yeah ok, if this isnt meant to be predicative - but then what's the point? is it a study for its own sake? I guess, what conclusions can we extract from it, and what things can we do with those conclusions?
I agree. Zone starts are anything but relative. If you play for a good team (like Letang), you get more opportunities to get +s over -s depending on your zone starts. You need to adjust for this between players on different teams.

Also for Qual comp. This should be adjusted relative to team, then compared to other players team adjusted qual comp. This gives you an idea if they got their results against 1st-4th liners. It makes a difference. Dominating 1st line opposition is much different from dominating 4th line.

A combination of these things gives you the big picture of what % was the player driving the results vs the situation they were placed in.

wgknestrick is offline   Reply With Quote
Old
08-17-2012, 11:01 AM
  #13
wgknestrick
Registered User
 
Join Date: Aug 2012
Posts: 2,146
vCash: 500
Quote:
Originally Posted by Sureves View Post
1. You can think it's not true if you want, but it is

2. Using save percentage ON (relative to team) is not a good idea because it's heavily driven by noise in the short run and will regress towards the mean in the long run. Fancy way to say it's mostly luck. Same with shooting percentage. The study done is performed to illustrate who had the better year - regardless of reasons (luck perhaps) - in terms of net positive production, not to determine who is the better overall player in the long-run.

As such, you don't need to account for sv percent on, or shooting percent on.
I disagree. SV% on when compared to qual comp ratings gives you an indication of the defensive ability of a defensive player. If they are matched up against 1st line opposition, (meaning they have 1-2 highest qual comp), that level of opposition is going to have an expected SH%. There is a relation between qual comp and SV% on a team level. You need to find out where your defender lies against the population. Is he posting higher than expected SV% relative to his qual comp or lower?

Defensive skaters can affect this by clearing out the crease and preventing close shots. The "noise" are usually just variables you may not have considered or have a difficult time measuring. Don't be so quick to throw logic out as soon as you run into results that seem to contain "noise". Try to adjust for situations/variables that may be contributing to it.

I obviously do not like this analysis for forwards' defensive ability, but for defenders, I think it has merit because they impact the number of high % shots.

wgknestrick is offline   Reply With Quote
Old
08-17-2012, 12:07 PM
  #14
seventieslord
Moderator
 
seventieslord's Avatar
 
Join Date: Mar 2006
Location: Regina, SK
Country: Canada
Posts: 24,750
vCash: 500
Does QoC really not matter? I find that hard to believe. Look at two players who have recently led the NHL in +/- despite low icetime: Marek Malik and Tom Preissing. This screams of dominant play in sheltered minutes. If you were to run numbers like this for these seasons for these players, would they not show up among the leaders in +/- relative to teammates? In reality though they were playing 3rd pairing minutes and weren't considered suitable opponents for the enemy's top tier forwards. QoC has to matter; I am pretty sure it is why they led the NHL in +/-. If QoC is available and properly applied it should show that these players were not actually their team's best (though perhaps overqualified to be 3rd pairing players based on their performance those particular seasons)

seventieslord is offline   Reply With Quote
Old
08-17-2012, 12:41 PM
  #15
Sureves
Registered User
 
Sureves's Avatar
 
Join Date: Sep 2008
Location: Ottawa
Posts: 7,131
vCash: 500
Quote:
Originally Posted by Verviticus View Post
zone starts aren't relative. can you please explain this more thoroughly?
Yeah sure, sorry, what I did above is using +/- per 60 relative to team, as well as +/- per 60 relative to 10-most common teammates, as such it is a relative stat.

As such, if we are going to consider zone starts %, it needs to be relative to the team he plays on. That way, if we are going to say for instance:

Player A Team A: +0.20 per 60 relative to team, +0.25 per 60 relative to teammates, 50% o-zone starts

versus

Player B Team B: +0.20 per 60 relative to team, +0.25 per 60 relative to teammates, 50% o-zone starts

We need to know the relative o-zone starts in order to see who was the better player between the two. Perhaps Team A is in the offensive zone 80% of the time and Team B is in the offensive zone only 20% of the time. The fact that Player B has 50% o-zone starts despite the fact that his team is almost never on the attack shows he is being used in an offensive role and only getting +0.25 per 60 relative to teammates. It should be easier for him to get +/- relative to team/teammates since he's being used in a more offensive role. While Player A is playing 50% o-zone starts despite the fact that his team is almost always on the attack and as such is being used in a defensive role, and yet he still manages to be +0.25 per 60.

Player A is better in this case (all else being equal of course) and that's why you need relative o-zone starts to use in conjunction with relative (to team/teammates) +/- and not nominal o-zone starts.

Quote:
yeah ok, if this isnt meant to be predicative - but then what's the point? is it a study for its own sake? I guess, what conclusions can we extract from it, and what things can we do with those conclusions?
Yeah fair enough, I made the study more as a means of showing why the Norris nominees were as they were this year, not as a means of being predictive in nature. I agree with you, if we wanted to take that approach you would need to normalize both shooting percentage and save percentage.

Sureves is offline   Reply With Quote
Old
08-17-2012, 12:50 PM
  #16
Sureves
Registered User
 
Sureves's Avatar
 
Join Date: Sep 2008
Location: Ottawa
Posts: 7,131
vCash: 500
Quote:
Originally Posted by seventieslord View Post
Does QoC really not matter? I find that hard to believe. Look at two players who have recently led the NHL in +/- despite low icetime: Marek Malik and Tom Preissing. This screams of dominant play in sheltered minutes. If you were to run numbers like this for these seasons for these players, would they not show up among the leaders in +/- relative to teammates? In reality though they were playing 3rd pairing minutes and weren't considered suitable opponents for the enemy's top tier forwards. QoC has to matter; I am pretty sure it is why they led the NHL in +/-. If QoC is available and properly applied it should show that these players were not actually their team's best (though perhaps overqualified to be 3rd pairing players based on their performance those particular seasons)
I've wondered this myself, and I personally reject the premise that QoC doesn't matter at all - despite the studies people have done on it.

I do however accept the fact that the difference over 82 games in QoC between any first pairing defenseman is so negligible that it doesn't warrant adjustment.

Sureves is offline   Reply With Quote
Old
08-17-2012, 01:04 PM
  #17
Sureves
Registered User
 
Sureves's Avatar
 
Join Date: Sep 2008
Location: Ottawa
Posts: 7,131
vCash: 500
Quote:
Originally Posted by wgknestrick View Post
I disagree. SV% on when compared to qual comp ratings gives you an indication of the defensive ability of a defensive player. If they are matched up against 1st line opposition, (meaning they have 1-2 highest qual comp), that level of opposition is going to have an expected SH%. There is a relation between qual comp and SV% on a team level. You need to find out where your defender lies against the population. Is he posting higher than expected SV% relative to his qual comp or lower?

Defensive skaters can affect this by clearing out the crease and preventing close shots. The "noise" are usually just variables you may not have considered or have a difficult time measuring. Don't be so quick to throw logic out as soon as you run into results that seem to contain "noise". Try to adjust for situations/variables that may be contributing to it.

I obviously do not like this analysis for forwards' defensive ability, but for defenders, I think it has merit because they impact the number of high % shots.
I'm not sure I entirely agree with that. Shooting percentage is widely considered to be erratic and not really a good representation of how good someone is at shooting the puck - at least when we look at only one year. Here were the 2012 top 30:


Curtis Glencross
Steven Stamkos
Jiri Hudler
Kyle Wellwood
Jordan Eberle
Nathan Horton
David Perron
Tyler Ennis
Martin Erat
Sergei Kostitsyn
Milan Lucic
Jordan Staal
Brad Marchand
Michael Ryder
Tyler Bozak
Maxime Talbot
Milan Michalek
Chris Kelly
Matt Moulson
David Desharnais
Valtteri Filppula
Scott Hartnell
Patrik Elias
David Krejci
Kyle Okposo
Nik Antropov
Teddy Purcell
Eric Nystrom
Jonathan Toews
Dainius Zubrus

Only 6 of 30 (bolded) scored 30 goals or more this year. I'm not sure it's fair to say those with higher shooting percentages (equivalently cause a goalie to have a low save percentage) are those who are of higher quality of competition.

I'm not of the opinion that in looking at two defenseman on the same team with exactly equal Corsi QoC, the one with the higher save percentage ON is the better defender. I think that's more based on luck than anything. I would much prefer to normalize the save percentages and then look at the GA relative to the team in question per 60. In other words, I'd look at shots against.

Again, this is kind of off-topic to what I posted specifically, but still an interesting discussion for sure.

Sureves is offline   Reply With Quote
Old
08-17-2012, 01:07 PM
  #18
barneyg
HFBoards Sponsor
 
Join Date: Apr 2007
Posts: 2,275
vCash: 500
Nice work btw. I just want to make two comments that I hope will be constructive.

Quote:
Originally Posted by Sureves View Post
Another thing to note with regards to QoC as I am quite sure this will come up in the posts in this thread, is that QoC plays almost no role in terms of a defenseman's advanced statistics over an 82 game period because they all face top competition, and even if one defenseman faces top competition more than the other, the difference in advanced statistics (ie. goals against, goals for, etc.) is so negligible that it isn't worth analysing (You can read about that here: http://nhlnumbers.com/2012/7/23/the-...of-competition).

As such, because these are all first pairing defenseman who face top competition on a regular basis, we can compare apples to apples here and no one should take issue with that and claim the "tougher competition" card. Please read the article if you are going to make such a post before posting to keep this thread clean.
I have read the article and I believe it is ill-suited for your purpose here (which is about ranking the league's best d-men among themselves). That article's conclusion is based on eyeballing a histogram for 3 Flyers players and showing that Nodl's expected Corsi (49.5%) is "only" 1.1% lower than Betts' (50.6%). Well guess what, the range in Corsi between the league's best and worst players is only ~12% (44-56) according to the article, so a 1% difference is significant! If d-men are uniformly distributed on Corsi and if there are 200 d-men in the league (6-7 regulars per team), that means the d-man who'd be 16th against good opposition could be 1st against mediocre opposition! If the uniformity assumption is wrong, he could leapfrog even more people.

Long story short: If you looked at the whole universe of d-men, then it may be true that QOC wouldn't really matter -- a guy ranked #50 out of 200 could really be 40-60 and it wouldn't matter. But if you're going to rank the league's top 10 d-men, QOC could make you conclude that a guy is ranked #2-3 when he should really be #10.

(I don't know how you can arbitrarily exclude Yandle on QOC and then make the claim that QOC is insignificant. I understand why you do it but to me it's just further evidence that QOC makes a difference, we're just not sure how to account for it.)

Quote:
Originally Posted by Sureves View Post
However, it should be noted that Kris Letang did not play nearly a full year this year, and it's possible that the games he played the team played better than they did on average (possibly as a result of the fact that he was playing in the first place?). Further analysis would be needed to properly account for this, but I will instead say that this is interesting for another study, but not in itself a clear indication that Letang is in fact as dominant as this indicates (though it potentially is).
"+/- relative to team" wouldn't pick up that effect as Behindthenet's "+/-OFF" only includes games for which the player actually played, i.e. if the Pens sucked while Letang was injured, it didn't affect his +/- relative to team because those games aren't considered. You can easily convince yourself of this by looking at d-men who have played very few games, their +/-OFF is all over the place. (ex: Sneep, Bortuzzo and Strait on the Pens)

barneyg is offline   Reply With Quote
Old
08-17-2012, 02:09 PM
  #19
Czech Your Math
Registered User
 
Czech Your Math's Avatar
 
Join Date: Jan 2006
Location: bohemia
Country: Czech_ Republic
Posts: 3,665
vCash: 500
If one completely filters out SV% while on ice (during SA), then doesn't that penalize d-men which prevent high quality shots? So a d-man who takes chances that either results in no shots or often in a high quality shot will appear better than one which prevents most high quality shots, but allows more low quality shots.

I understand the luck component in the shorter term, so maybe it would be better to see results both ways. In a longer term study, one could get an average of SV% on/off for each d-man and adjust based on that, right?

Czech Your Math is offline   Reply With Quote
Old
08-17-2012, 04:21 PM
  #20
Sureves
Registered User
 
Sureves's Avatar
 
Join Date: Sep 2008
Location: Ottawa
Posts: 7,131
vCash: 500
Quote:
Originally Posted by Czech Your Math View Post
If one completely filters out SV% while on ice (during SA), then doesn't that penalize d-men which prevent high quality shots? So a d-man who takes chances that either results in no shots or often in a high quality shot will appear better than one which prevents most high quality shots, but allows more low quality shots.

I understand the luck component in the shorter term, so maybe it would be better to see results both ways. In a longer term study, one could get an average of SV% on/off for each d-man and adjust based on that, right?
To your first paragraph, that's right, and I definitely understand the criticisms you have for it because they are warranted. If a defenseman is doing bone-headed plays on a regular basis and causing breakaways for the other team, naturally that would cause them to have a lower save percentage ON and that would be representative of how good (or bad) they are defensively.

Again of course the problem comes from the fact that it's too difficult in the short run (ie. a season) to see if it's just luck or their fault. If you looked at a wider sample of seasons it would clarify the result a lot and would definitely be interesting to do.

Maybe something to consider when I have some more time.

Sureves is offline   Reply With Quote
Old
08-17-2012, 04:31 PM
  #21
Sureves
Registered User
 
Sureves's Avatar
 
Join Date: Sep 2008
Location: Ottawa
Posts: 7,131
vCash: 500
Quote:
Originally Posted by barneyg View Post
Nice work btw. I just want to make two comments that I hope will be constructive.



I have read the article and I believe it is ill-suited for your purpose here (which is about ranking the league's best d-men among themselves). That article's conclusion is based on eyeballing a histogram for 3 Flyers players and showing that Nodl's expected Corsi (49.5%) is "only" 1.1% lower than Betts' (50.6%). Well guess what, the range in Corsi between the league's best and worst players is only ~12% (44-56) according to the article, so a 1% difference is significant! If d-men are uniformly distributed on Corsi and if there are 200 d-men in the league (6-7 regulars per team), that means the d-man who'd be 16th against good opposition could be 1st against mediocre opposition! If the uniformity assumption is wrong, he could leapfrog even more people.

Long story short: If you looked at the whole universe of d-men, then it may be true that QOC wouldn't really matter -- a guy ranked #50 out of 200 could really be 40-60 and it wouldn't matter. But if you're going to rank the league's top 10 d-men, QOC could make you conclude that a guy is ranked #2-3 when he should really be #10.

(I don't know how you can arbitrarily exclude Yandle on QOC and then make the claim that QOC is insignificant. I understand why you do it but to me it's just further evidence that QOC makes a difference, we're just not sure how to account for it.)



"+/- relative to team" wouldn't pick up that effect as Behindthenet's "+/-OFF" only includes games for which the player actually played, i.e. if the Pens sucked while Letang was injured, it didn't affect his +/- relative to team because those games aren't considered. You can easily convince yourself of this by looking at d-men who have played very few games, their +/-OFF is all over the place. (ex: Sneep, Bortuzzo and Strait on the Pens)
Not sure I entirely understand the first part of your post (sorry it's been a long week ) but thanks for sharing about the +/- relative to team, didn't know it was already adjusted to only include games you played in...makes the Letang results very interesting (though as you said, QoC is something to consider).

Sureves is offline   Reply With Quote
Old
08-17-2012, 06:22 PM
  #22
Verviticus
Registered User
 
Join Date: Jul 2010
Posts: 4,989
vCash: 500
Quote:
Originally Posted by Sureves View Post
Yeah sure, sorry, what I did above is using +/- per 60 relative to team, as well as +/- per 60 relative to 10-most common teammates, as such it is a relative stat.

As such, if we are going to consider zone starts %, it needs to be relative to the team he plays on. That way, if we are going to say for instance:

Player A Team A: +0.20 per 60 relative to team, +0.25 per 60 relative to teammates, 50% o-zone starts

versus

Player B Team B: +0.20 per 60 relative to team, +0.25 per 60 relative to teammates, 50% o-zone starts

We need to know the relative o-zone starts in order to see who was the better player between the two. Perhaps Team A is in the offensive zone 80% of the time and Team B is in the offensive zone only 20% of the time. The fact that Player B has 50% o-zone starts despite the fact that his team is almost never on the attack shows he is being used in an offensive role and only getting +0.25 per 60 relative to teammates. It should be easier for him to get +/- relative to team/teammates since he's being used in a more offensive role. While Player A is playing 50% o-zone starts despite the fact that his team is almost always on the attack and as such is being used in a defensive role, and yet he still manages to be +0.25 per 60.

Player A is better in this case (all else being equal of course) and that's why you need relative o-zone starts to use in conjunction with relative (to team/teammates) +/- and not nominal o-zone starts.



Yeah fair enough, I made the study more as a means of showing why the Norris nominees were as they were this year, not as a means of being predictive in nature. I agree with you, if we wanted to take that approach you would need to normalize both shooting percentage and save percentage.
fair enough. thanks!

Verviticus is offline   Reply With Quote
Old
08-17-2012, 08:13 PM
  #23
barneyg
HFBoards Sponsor
 
Join Date: Apr 2007
Posts: 2,275
vCash: 500
Quote:
Originally Posted by Sureves View Post
Not sure I entirely understand the first part of your post (sorry it's been a long week ) but thanks for sharing about the +/- relative to team, didn't know it was already adjusted to only include games you played in...makes the Letang results very interesting (though as you said, QoC is something to consider).
My point is that the "QOC doesn't matter" conclusion can only be valid if you're looking to split players into broad skill tiers, say: great, average, replacement level. Letang, Keith, Karlsson, etc are always going to show up in the "great" category no matter their QOC -- the uncorrected bias isn't going to be enough to move them down one category. But it's likely to be enough to move a guy from #3 to #10 or vice-versa. So if your goal is to compare the top d-men with each other, you can't say "QOC doesn't matter" and ignore it.

That said, IMO the zone start adjustment is likely to be more important than the QOC adjustment.

barneyg is offline   Reply With Quote
Old
08-17-2012, 09:32 PM
  #24
overpass
Registered User
 
Join Date: Jun 2007
Posts: 3,601
vCash: 500
Quote:
Originally Posted by barneyg View Post
My point is that the "QOC doesn't matter" conclusion can only be valid if you're looking to split players into broad skill tiers, say: great, average, replacement level. Letang, Keith, Karlsson, etc are always going to show up in the "great" category no matter their QOC -- the uncorrected bias isn't going to be enough to move them down one category. But it's likely to be enough to move a guy from #3 to #10 or vice-versa. So if your goal is to compare the top d-men with each other, you can't say "QOC doesn't matter" and ignore it.

That said, IMO the zone start adjustment is likely to be more important than the QOC adjustment.
Zone start vs QoC depends on whether you are using shot-based or goal-based metrics. QoC is relatively more important for goal-based metrics like plus-minus (because top players have better SH%) and zone start info is relatively more important for shot-based metrics like Corsi (because shots off an o-zone faceoff tend to be low %).

overpass is offline   Reply With Quote
Old
08-17-2012, 10:27 PM
  #25
wgknestrick
Registered User
 
Join Date: Aug 2012
Posts: 2,146
vCash: 500
Quote:
Originally Posted by Sureves View Post
I'm not sure I entirely agree with that. Shooting percentage is widely considered to be erratic and not really a good representation of how good someone is at shooting the puck - at least when we look at only one year. Here were the 2012 top 30:


Curtis Glencross
Steven Stamkos
Jiri Hudler
Kyle Wellwood
Jordan Eberle
Nathan Horton
David Perron
Tyler Ennis
Martin Erat
Sergei Kostitsyn
Milan Lucic
Jordan Staal
Brad Marchand
Michael Ryder
Tyler Bozak
Maxime Talbot
Milan Michalek
Chris Kelly
Matt Moulson
David Desharnais
Valtteri Filppula
Scott Hartnell
Patrik Elias
David Krejci
Kyle Okposo
Nik Antropov
Teddy Purcell
Eric Nystrom
Jonathan Toews
Dainius Zubrus

Only 6 of 30 (bolded) scored 30 goals or more this year. I'm not sure it's fair to say those with higher shooting percentages (equivalently cause a goalie to have a low save percentage) are those who are of higher quality of competition.

I'm not of the opinion that in looking at two defenseman on the same team with exactly equal Corsi QoC, the one with the higher save percentage ON is the better defender. I think that's more based on luck than anything. I would much prefer to normalize the save percentages and then look at the GA relative to the team in question per 60. In other words, I'd look at shots against.

Again, this is kind of off-topic to what I posted specifically, but still an interesting discussion for sure.
What does a defender's SV% on ice have to do with these forwards' SH%?

wgknestrick is offline   Reply With Quote
Reply

Forum Jump


Bookmarks

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off



All times are GMT -5. The time now is 06:14 AM.

monitoring_string = "e4251c93e2ba248d29da988d93bf5144"
Contact Us - HFBoards - Archive - Privacy Statement - Terms of Use - Advertise - Top - AdChoices

vBulletin Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
HFBoards.com is a property of CraveOnline Media, LLC, an Evolve Media, LLC company. 2014 All Rights Reserved.