Colorado Avalanche's success - will it last?

[Bear of Bad News]

Puck Daddy article on the Avalanche's defiance of advanced analytics thus far in 2013-14:

http://sports.yahoo.com/blogs/puck-...unsustainable--trending-topics-205937984.html

Click to read more...

 

[YoSoyLalo]

There's always an exception to the rule.

I don't think the Avs will win a cup, but I do think they could go fairly deep this year. I think a big part of their success, something that has helped balance out the lack of good advanced numbers, is Varlamov's goaltending. He should be a Hart candidate, IMO. That, plus timely scoring.

 

[SnowblindNYR]

Yeah out of the top teams in the west they seem to be the weakest. They've been getting a lot of clutch goals late to tie games which is partially a product of their offensive firepower and pressure and of Roy's unconventional coaching (pulling the goalie early) but that kind of stuff is unsustainable. The Avs are considered a bigger contender than the Rangers but I'm fairly confident that if both teams made the finals (both would be upsets at this point) the Rangers would win, can't say I'd be confident against any other Western hopeful like that.

 

[Ishdul]

They should be favoured over Minnesota; near equal possession stats but with home ice and a better goaltending situation, although I guess Duchene is likely out which adds some questions. And they should be underdogs vs. whoever comes out of St. Louis/Chicago, but I don't think that should surprise anyone.

My guess is they will become more sustainably good in the coming years, where a guy like MacKinnon will go from currently being a bit of a possession black hole as a talented but still green rookie into a dominant possession player and many of their other players will improve with experience, there cap situation is very solid with no obvious bad contracts and lots of room to add salary and I think some willingness to spend. MacKinnon is also the type of player that should reliably outperform his possession stats, since he who will generate a ton of odd-man rushes with his phenomenal skating and has had ultra high shooting rates at every level IIRC. Varlamov likely won't match the heights of this season and I'm still wary about him getting injured but I don't think this is far off from his talent level that he showed in Washington.

 

[Ceremony]

I don't post in this forum so I don't know what the standard etiquette for topics like this is, but the main reason I don't really care about advanced stats (aside from one time I looked up how valuable Jay McClement was to the Avs on behindthenet and got a headache) is because of the smugness of some people who treat them as the only thing that can truly tell you how good a team is, and mock anyone who doesn't immediately agree with them.

Why do so many articles about advanced stats have to sound so utterly condescending?

That what they do to create the appearance of luck is, in fact, a repeatable skill that they're going to be able to churn out forever and ever under this system and with these players.

That what they do is somehow different from the other teams, those that have already failed in much the same way the numbers say they will as well, and therefore it doesn't matter how badly they get outshot or how high their shooting percentage is or how much larger their goaltender inflates his save percentage over his career average.

They're just going to keep winning. Because they're different.

Of course, there's no chance in the future of any of the players changing, or improving. The Avalanche will just keep going with their assortment of #6 defensemen.

A lot of their success this season has been attributed to their new coach, Patrick Roy, because hey he's a legendary goaltender and well-liked figure, and the team immediately turned around under him. It's like none of them have any memory whatsoever of the Joe Sacco-led trip to the playoffs, which itself was made possible only by an outlier of a wonderful 50ish-game goaltending performance from Craig Anderson.

The season, he found to his dismay, is a little longer than 50 or so games, and the Avs were ground from a top team with a little ways to go down to – guess what! – a second-place finish in the division, and were summarily bounced in six playoff games by a deeply superior San Jose Sharks side.

Does no one in Colorado remember this having happened? What about it is different, other than Semyon Varlamov having kept up the game-breaking goaltending for far longer than most would have expected?

Wow! This article seems to come from a well-informed source! The team is exactly the same as it was four years ago even though there's only five players still remaining from then (two of whom have improved dramatically in O'Reilly in Duchene, one of whom is an irrelevance in Wilson). They play exactly the same way and have all the same strengths, weaknesses and styles of play.

It gets worse when you see an article as bad as this making a bunch of utterly redundant comparisons and assumptions. "They're only winning because of how good the goalie is!" Okay. isn't the goalie part of the team? Does this mean all the teams that have won a Cup with their goalie getting the Conn Smythe relied on him too much? "They're going to regress eventually, they can't play the way they are and keep winning!" Roy has said as much as this many times this year, and I'm sure he knows better than anyone the deficiencies of his team (as I would expect of any coach).

I also think it's patronising to dismiss coming from behind in games or games with a great defensive effort to win as lucky. Look at game 1 for the Avs in the playoffs, they're down by two quick goals the previous period to start the third, a team with very little collective playoff experience, and they don't give up. They keep pushing at it, playing their game, and then tie and then win it. To dismiss that and the results all year as "luck" in the face of some of the adversity the team faced just makes my blood boil, because it completely ignores why the team is successful while contradicting the stats that some people cling to as the only thing that can tell them what teams will and what teams won't succeed.

 

[Strong Island]

I view them as a team with a terrible bottom six, boosted by incredible goaltending and insane shooting luck by their defensemen (seriously 7.6% shooting from d-men?).

The bottom six is an easy fix, the goaltending could easily be real, and the development of their young forwards should make up for the regression in defensive goal contribution.

As long as management recognizes that there are serious flaws that are fixable and don't become attached to garbage bottom sixers, they could definitely be a force going forward.

 

[The Mighty Duck Man]

I don't post in this forum so I don't know what the standard etiquette for topics like this is, but the main reason I don't really care about advanced stats (aside from one time I looked up how valuable Jay McClement was to the Avs on behindthenet and got a headache) is because of the smugness of some people who treat them as the only thing that can truly tell you how good a team is, and mock anyone who doesn't immediately agree with them.

Why do so many articles about advanced stats have to sound so utterly condescending?



Of course, there's no chance in the future of any of the players changing, or improving. The Avalanche will just keep going with their assortment of #6 defensemen.



Wow! This article seems to come from a well-informed source! The team is exactly the same as it was four years ago even though there's only five players still remaining from then (two of whom have improved dramatically in O'Reilly in Duchene, one of whom is an irrelevance in Wilson). They play exactly the same way and have all the same strengths, weaknesses and styles of play.

It gets worse when you see an article as bad as this making a bunch of utterly redundant comparisons and assumptions. "They're only winning because of how good the goalie is!" Okay. isn't the goalie part of the team? Does this mean all the teams that have won a Cup with their goalie getting the Conn Smythe relied on him too much? "They're going to regress eventually, they can't play the way they are and keep winning!" Roy has said as much as this many times this year, and I'm sure he knows better than anyone the deficiencies of his team (as I would expect of any coach).

I also think it's patronising to dismiss coming from behind in games or games with a great defensive effort to win as lucky. Look at game 1 for the Avs in the playoffs, they're down by two quick goals the previous period to start the third, a team with very little collective playoff experience, and they don't give up. They keep pushing at it, playing their game, and then tie and then win it. To dismiss that and the results all year as "luck" in the face of some of the adversity the team faced just makes my blood boil, because it completely ignores why the team is successful while contradicting the stats that some people cling to as the only thing that can tell them what teams will and what teams won't succeed.

Honestly, I think half the people who claim to be part of the "analytics community" don't even understand most of the stats and are just in it to feel like they're better than others. Maybe that's pushing it, but it does explain why some of the worst offenders(for being condescending) also don't even seem to fully understand what they're rattling off. This guy seems like one of them. Sure, the Avs are defying the stats, but aside from listing the corsi numbers from their season series with the Hawks, he literally doesn't talk about numbers at all, just making stupid comparisons to past Colorado teams and the Leafs.

Does Puck Daddy just let anyone write for them now? Not that they were ever great, but still, wow.

 

[Bonzai12]

Stats are just a tool in the toolbox.

 

[Gigantor The Goalie]

Stats are just a tool in the toolbox.

In this case trying to use a hammer to paint a wall

 

[Bear of Bad News]

In this case trying to use a hammer to paint a wall

That's not a fair characterization of the statistics that the author is bringing up (I'm an Avalanche fan, by the way).

The metrics have merit, and it's a valid question. You can disagree with how the author portrayed it, but it's a valid question.

 

[Steerpike]

I view them as a team with a terrible bottom six.

The Avs are currently missing their top forward, a top six forward, and their third line center. Should we really expect their bottom six not to look pretty rough?

boosted by incredible goaltending and insane shooting luck by their defensemen (seriously 7.6% shooting from d-men?).

Actually watch a game and you'll see that most of the goals from the Dmen have come not from blue line slappers but from Barrie/Holden/Benoit crashing really deep into the zone. The chart on Nick Holden seems to have assigned the goals to random shots, but about eight of his goals have come from sneaking in and tipping in a cross crease pass.

 

[Freudian]

I wonder what Roy using man on man defense does to shot quality for the other team. Roy has several times said he would prefer not to allow as many shots but also said several times after being heavily outshot that as far as scoring chances go, they are satisfied. Roy doesn't seem that into advanced stats but the team is tracking scoring chances and seem to put a lot of weight in them. Roy has recounted scoring chance numbers quite a bit in post game press conferences.

Compared to the Sacco teams (which allowed fewer shots) Avs opponents find themselves wide open right in the slot less because they always are supposed to have an Avs player tracking them.

Any realistic Avs fans realize the team has been punching above their weight class this season and has been a bit lucky when it comes to several things (bounces, schedule, injuries). But those that only evaluate the team from the corsi/fenwick numbers seem to miss why the team is doing well.

 

[Strong Island]

The Avs are currently missing their top forward, a top six forward, and their third line center. Should we really expect their bottom six not to look pretty rough?



Actually watch a game and you'll see that most of the goals from the Dmen have come not from blue line slappers but from Barrie/Holden/Benoit crashing really deep into the zone. The chart on Nick Holden seems to have assigned the goals to random shots, but about eight of his goals have come from sneaking in and tipping in a cross crease pass.

Insults aren't necessary.

I wouldn't argue against the notion that most of their goals resulted from crashing in and taking close in shots for goals. I would argue that this is not a repeatable skill season to season and should not be relied upon for a team's continued success. Looking at his average shot distance and *gasp* watching him play, it seems like Barrie has a true skill which helps decrease his average shot distance. The others though? I think it's pretty safe to say Nick Holden isn't going to shoot 15% over a 50 game stretch for the rest of his career.

 

[OilerPensfan97]

Not sure when they'll win the cup, but I think it will be a very long time until they miss the playoffs.

 

[Mr Positive]

I wonder what Roy using man on man defense does to shot quality for the other team. Roy has several times said he would prefer not to allow as many shots but also said several times after being heavily outshot that as far as scoring chances go, they are satisfied. Roy doesn't seem that into advanced stats but the team is tracking scoring chances and seem to put a lot of weight in them. Roy has recounted scoring chance numbers quite a bit in post game press conferences.

Compared to the Sacco teams (which allowed fewer shots) Avs opponents find themselves wide open right in the slot less because they always are supposed to have an Avs player tracking them.

Any realistic Avs fans realize the team has been punching above their weight class this season and has been a bit lucky when it comes to several things (bounces, schedule, injuries). But those that only evaluate the team from the corsi/fenwick numbers seem to miss why the team is doing well.

I agree to an extent but remember that this is what Leafs fans were saying as well. Good for the Avs for sustaining this pace but there may be a weakness there. It may only be a matter of time before it is exposed and exploited.

 

[Bear of Bad News]

I agree to an extent but remember that this is what Leafs fans were saying as well. Good for the Avs for sustaining this pace but there may be a weakness there. It may only be a matter of time before it is exposed and exploited.

I would also expect the team to get older and have better advanced metrics as seasons go on.

Their current success may be unsustainable given their advanced metrics, but it may not have to be.

 

[Hennessy]

Wow! This article seems to come from a well-informed source! The team is exactly the same as it was four years ago even though there's only five players still remaining from then (two of whom have improved dramatically in O'Reilly in Duchene, one of whom is an irrelevance in Wilson). They play exactly the same way and have all the same strengths, weaknesses and styles of play.

It's Lambert. He gets his clicks by trolling. Or "generating controversy", excuse me. He's not going to be reasonable as that might undermine the Utter Authority his shtick depends on. At any rate, I find it funny that he's the Puck Daddy metrics guy as his arguments tend to fall straight into emotion faster than any of their other writers. Which is something brought up in this thread about how some guys hide behind numbers. Lambert would be one of those. And which by no means indicates metrics are crap or that Colorado didn't get some luck this season. But Lambert wields them like a sharp stick to whichever beehive he thinks he can stir up.

 

[kitsel]

I'm a huge proponent of advanced stats in games like baseball where even metric is trackable and each play is an isolated case where everything can be taken into account. But Hockey simply isn't.

I think the advanced group is as quick to dismiss that their formulas/stats may be lacking certain variables as the other crowd is of dismissing any notion that they might be getting lucky or playing over their heads. When teams like Anaheim defy the statistics with regularity it shows me that things are being neglected. And those things are pretty obvious - yes it's "supposed" to balance out, but goalie quality, shot location, shot quality, player quality (are we really going to say a shot from Perry is the same as a blue line shot from some random 4th liner)... none of these things are taken into account. It's great in theory to say that "over a full season it all balances out" but I'm not sure it does. I'd love to see someone go through the outliers and see if everything is within 2 standard deviations (in # of outliers and HOW far of an outlier they are).

One thing that has never made sense to me is that if these statistics are so volatile and randomness based that a team like the Avalanche can be a *huge* outlier over 82 games, why is it impossible for them to continue to be an outlier in a 7 game series? Why is an 82 game regular season insufficient to "validate" anything but a 7 game playoff series is? Undoubtedly if the Avalanche lose to the Hawks or blues this postseason it will be VALIDATION!!!!!!! for the stats crew. But why is that any more important than the entirety of the regular season or the almost-as-long regular season series? If Colorado flew in the face of corsi in the regular season in a "cartoonish" way, why isn't that even POSSIBLE in a 7 game series? I'm a huge proponent of the advanced statistics in baseball but they are just so much more concrete. The hockey advanced statistics proponents seem very quick to throw numbers out for people so supposedly interested in the numbers.

The individual player corsi and fenwick ratings are just as concerning to me. I'm going to copy/paste a post I made earlier today -

"Was looking through the advanced statistics for the Avs when I found this little tid-bit.

According to iCorsi/60, in a sample size of over 750 minutes for each of them, Jamie McGinn was "better" than
RoR, Stastny, Mitchell, Talbot, PaP, Barrie, EJ, Holden, Benoit, Sarich, Guenin, and Hejda, in that order.

McGinn is considered the Avalanche's fourth best player by iCorsi/60

According to iFenwick/60, in that same sample size, Jamie McGinn was "better" than
Landy, Stastny, RoR, PaP, Talbot, Mitchell, Barrie, EJ, Holden, Benoit, Sarich, Hejda, and Guenin in that order.

McGinn is considered the Avalanche's third best player by iFenwick/60

Yet no one in their right mind would EVER pick McGinn over RoR or Stastny. Isn't this basis enough to say that Corsi and Fenwick are flawed?

Now, don't get me wrong here. I am a huge fan of advanced statistics in baseball because every minutia of every play is able to be taken into account. It's a slow, play by play game where everything can be accounted for. Hockey is so quick and advanced statistics are so new that I just frankly think they're dangerous to take too seriously. I hope they're improved as time goes on but I can't help but feel like massive, large sample size failures like this are indicative of variables not being taken into account more than they are outliers or simple luck. Teams like the Ducks defy the advanced statistics frequently and to me that screams inadequate information more than it does regression to the mean. Obviously, having a high Corsi/Fenwick/sustainable PDO are a good thing but I really believe there's stuff going on that we aren't taking into account that are accounting for a lot of the outliers.

Am I missing something here or are these stats as egregious as they appear?

http://stats.hockeyanalysis.com/rat...rs&minutes=750&disp=1&sort=HARTp&sortdir=DESC "

I'm not saying that advanced statistics are useless, or that the Avs WILL NOT take a step back next year. Obviously possession is great to have and more shots are great but aren't all those factors I've pointed out important as well? Isn't it at least possible that there's more going on than the current advanced statistics are able to analyze? It isn't a show of weakness to admit that the rather crude advanced statistics do not CURRENTLY show the whole picture. It doesn't discount them as a whole just because we haven't figured out every little thing.

But when you've got sample sizes of 1000 minutes telling you that Jamie Mc-freaking-Ginn outperformed Ryan O'Reilly and Paul Stastny, isn't that a call to reevaluate these statistics? Would any of you with complete faith in the current statistics actually trust the numbers and choose McGinn over those two given the same salary? By the way, Fenwick and Corsi had McGinn at #2 on the avs last year, ahead of even Matt Duchene. Purely by the stats he's better than Duchene, Stastny, or O'Reilly. Do those of you who trust the stats trust them enough to pick McGinn over those 3? Because one thing I've learned as a baseball stat junkie is that you can't pick and choose to ignore certain stats. You take them as they are or you admit that there are flaws in the system and try to fix them. McGinn outperforming RoR and Stastny 3 seasons in a row, the entirety of his time with the avs, is beyond the point of "luck" or "outlier."

 

[Bear of Bad News]

I think the advanced group is as quick to dismiss that their formulas/stats may be lacking certain variables as the other crowd is of dismissing any notion that they might be getting lucky or playing over their heads.

I don't know anyone in the "advanced group" that thinks that their methodology is without flaws.

Anyone who should be taken seriously, at least.

 

[kitsel]

I don't know anyone in the "advanced group" that thinks that their methodology is without flaws.

Anyone who should be taken seriously, at least.

Fair enough - actually from what I've noticed the people doing this are people who just like to be part of an "elite" group. I'm not sure they entirely understand the advanced statistics, they just like trumpeting them and appearing to be "in the know" ahead of time.

Bit of an unfair thing for me to say, in retrospect.

 

[Bear of Bad News]

Fair enough - actually from what I've noticed the people doing this are people who just like to be part of an "elite" group. I'm not sure they entirely understand the advanced statistics, they just like trumpeting them and appearing to be "in the know" ahead of time.

Bit of an unfair thing for me to say, in retrospect.

There are a few groups - those who want to be cool, and those who want to figure out as much as they can.

One of the benefits of knowing that we'll almost surely never get to the finish line is that we can revel in the fact that we can always make progress. There's always something new to learn.

 

[Freudian]

I would also expect the team to get older and have better advanced metrics as seasons go on.

Their current success may be unsustainable given their advanced metrics, but it may not have to be.

The hockey world resets every July. New players, new coaches, players get older for better and for worse. It's not a complete reset, but a big enough for last years stats (advanced and regular) to lose a bit of whatever relevance they had.

 

[Hammer Time]

I'm a huge proponent of advanced stats in games like baseball where even metric is trackable and each play is an isolated case where everything can be taken into account. But Hockey simply isn't.

I think the advanced group is as quick to dismiss that their formulas/stats may be lacking certain variables as the other crowd is of dismissing any notion that they might be getting lucky or playing over their heads. When teams like Anaheim defy the statistics with regularity it shows me that things are being neglected. And those things are pretty obvious - yes it's "supposed" to balance out, but goalie quality, shot location, shot quality, player quality (are we really going to say a shot from Perry is the same as a blue line shot from some random 4th liner)... none of these things are taken into account. It's great in theory to say that "over a full season it all balances out" but I'm not sure it does. I'd love to see someone go through the outliers and see if everything is within 2 standard deviations (in # of outliers and HOW far of an outlier they are).

One thing that has never made sense to me is that if these statistics are so volatile and randomness based that a team like the Avalanche can be a *huge* outlier over 82 games, why is it impossible for them to continue to be an outlier in a 7 game series? Why is an 82 game regular season insufficient to "validate" anything but a 7 game playoff series is? Undoubtedly if the Avalanche lose to the Hawks or blues this postseason it will be VALIDATION!!!!!!! for the stats crew. But why is that any more important than the entirety of the regular season or the almost-as-long regular season series? If Colorado flew in the face of corsi in the regular season in a "cartoonish" way, why isn't that even POSSIBLE in a 7 game series? I'm a huge proponent of the advanced statistics in baseball but they are just so much more concrete. The hockey advanced statistics proponents seem very quick to throw numbers out for people so supposedly interested in the numbers.

The individual player corsi and fenwick ratings are just as concerning to me. I'm going to copy/paste a post I made earlier today -

"Was looking through the advanced statistics for the Avs when I found this little tid-bit.

According to iCorsi/60, in a sample size of over 750 minutes for each of them, Jamie McGinn was "better" than
RoR, Stastny, Mitchell, Talbot, PaP, Barrie, EJ, Holden, Benoit, Sarich, Guenin, and Hejda, in that order.

McGinn is considered the Avalanche's fourth best player by iCorsi/60

According to iFenwick/60, in that same sample size, Jamie McGinn was "better" than
Landy, Stastny, RoR, PaP, Talbot, Mitchell, Barrie, EJ, Holden, Benoit, Sarich, Hejda, and Guenin in that order.

McGinn is considered the Avalanche's third best player by iFenwick/60

Yet no one in their right mind would EVER pick McGinn over RoR or Stastny. Isn't this basis enough to say that Corsi and Fenwick are flawed?

Now, don't get me wrong here. I am a huge fan of advanced statistics in baseball because every minutia of every play is able to be taken into account. It's a slow, play by play game where everything can be accounted for. Hockey is so quick and advanced statistics are so new that I just frankly think they're dangerous to take too seriously. I hope they're improved as time goes on but I can't help but feel like massive, large sample size failures like this are indicative of variables not being taken into account more than they are outliers or simple luck. Teams like the Ducks defy the advanced statistics frequently and to me that screams inadequate information more than it does regression to the mean. Obviously, having a high Corsi/Fenwick/sustainable PDO are a good thing but I really believe there's stuff going on that we aren't taking into account that are accounting for a lot of the outliers.

Am I missing something here or are these stats as egregious as they appear?

http://stats.hockeyanalysis.com/rat...rs&minutes=750&disp=1&sort=HARTp&sortdir=DESC "

I'm not saying that advanced statistics are useless, or that the Avs WILL NOT take a step back next year. Obviously possession is great to have and more shots are great but aren't all those factors I've pointed out important as well? Isn't it at least possible that there's more going on than the current advanced statistics are able to analyze? It isn't a show of weakness to admit that the rather crude advanced statistics do not CURRENTLY show the whole picture. It doesn't discount them as a whole just because we haven't figured out every little thing.

But when you've got sample sizes of 1000 minutes telling you that Jamie Mc-freaking-Ginn outperformed Ryan O'Reilly and Paul Stastny, isn't that a call to reevaluate these statistics? Would any of you with complete faith in the current statistics actually trust the numbers and choose McGinn over those two given the same salary? By the way, Fenwick and Corsi had McGinn at #2 on the avs last year, ahead of even Matt Duchene. Purely by the stats he's better than Duchene, Stastny, or O'Reilly. Do those of you who trust the stats trust them enough to pick McGinn over those 3? Because one thing I've learned as a baseball stat junkie is that you can't pick and choose to ignore certain stats. You take them as they are or you admit that there are flaws in the system and try to fix them. McGinn outperforming RoR and Stastny 3 seasons in a row, the entirety of his time with the avs, is beyond the point of "luck" or "outlier."

Re: Individual Stats

First of all, "iCorsi" is not to be confused with "On-ice Corsi", the statistic normally referred to as Corsi for short. HockeyAnalysis' "iCorsi" stat means "number of shot attempts taken by an individual player per 60 minutes of 5v5 ice time". "On-ice Corsi" is "number of shot attempts for minus shot attempts against taken by anyone while an individual is on the ice at 5v5", so basically +/- but with shots instead of goals.

Second, there are too many variables in a hockey game for a player's On-ice Corsi to be reliable taken by iself. A player's Corsi can be inflated by having good linemates, or if they are sheltered by their coach by not having to play against opponent top lines.

There exist other statistics such as Quality of Competition (QoC) - of which there are several variants - which attempt to quantify a player's role (i.e. are they playing against opponent top lines or are they getting sheltered shifts against opponent 3rd and 4th lines?), and Quality of Teammates (QoT), which attempts to quantify how good a player's linemates are. I personally don't find QoT particularly useful, because you can just look at boxscores to find out who a player's usual linemates are. Usually if you watch a team enough, you already know who has high QoC, but for teams you're unfamiliar with I find it does a good job of identifying the players who have a tougher role.

Time on ice is also a good number to look at - the more shifts you have, the harder it is to maintain a good Corsi.

http://www.behindthenet.ca/nhl_stat...+18+19+20+21+22+29+30+32+33+34+35+36+37+38+63

Here is the Avalanche's On-ice Corsi for this season. The players you expect to be at the top are at the top, except for Barrie-Holden who have a much lower QoC than the rest.

Re: Team Stats

Of course Corsi and Fenwick are incomplete. By definition, they only count what the skaters do - not the goalies. By definition, special teams don't count. So if you're a team with good special teams or goaltending, you will play better than your Corsi/Fenwick indicates. The theory is that, because goaltending, shooting talent, and special teams are so volatile (i.e. Smith, Mason, Price, Hiller, and Anderson among others have had both amazing and terrible seasons, Clarkson's shooting percentage has fluctuated between 5 and 15% the last few seasons, and few teams sustain a good PP or PK from season to season), it's best to just leave them out of the statistic itself and throw them into the catch-all category of "luck". It's then up to the person reading the stats to identify which PDO outliers are actually "luck" and which ones might be somewhat sustainable.

The expectation is that a human looking at the numbers will say "the Avalanche have a poor Corsi and a high PDO, but Varlamov just got a goalie coach this season so he might actually have become an elite goalie, and that PDO isn't so lucky after all". Or "Stamkos has a really high PDO, but he's just a great, talented shooter and he can sustainably produce higher-quality scoring chances than the average player". Most of the people that developed these advanced stats know their limitations. The problem is that they don't do a very good job of educating the general public about their stats, so you get articles on Puck Daddy that don't seem to understand Corsi and PDO very well and draw conclusions that are beyond what the data suggest.

 

[kitsel]

Re: Individual Stats

First of all, "iCorsi" is not to be confused with "On-ice Corsi", the statistic normally referred to as Corsi for short. HockeyAnalysis' "iCorsi" stat means "number of shot attempts taken by an individual player per 60 minutes of 5v5 ice time". "On-ice Corsi" is "number of shot attempts for minus shot attempts against taken by anyone while an individual is on the ice at 5v5", so basically +/- but with shots instead of goals.

Second, there are too many variables in a hockey game for a player's On-ice Corsi to be reliable taken by iself. A player's Corsi can be inflated by having good linemates, or if they are sheltered by their coach by not having to play against opponent top lines.

There exist other statistics such as Quality of Competition (QoC) - of which there are several variants - which attempt to quantify a player's role (i.e. are they playing against opponent top lines or are they getting sheltered shifts against opponent 3rd and 4th lines?), and Quality of Teammates (QoT), which attempts to quantify how good a player's linemates are. I personally don't find QoT particularly useful, because you can just look at boxscores to find out who a player's usual linemates are. Usually if you watch a team enough, you already know who has high QoC, but for teams you're unfamiliar with I find it does a good job of identifying the players who have a tougher role.

Time on ice is also a good number to look at - the more shifts you have, the harder it is to maintain a good Corsi.

http://www.behindthenet.ca/nhl_stat...+18+19+20+21+22+29+30+32+33+34+35+36+37+38+63

Here is the Avalanche's On-ice Corsi for this season. The players you expect to be at the top are at the top, except for Barrie-Holden who have a much lower QoC than the rest.

Re: Team Stats

Of course Corsi and Fenwick are incomplete. By definition, they only count what the skaters do - not the goalies. By definition, special teams don't count. So if you're a team with good special teams or goaltending, you will play better than your Corsi/Fenwick indicates. The theory is that, because goaltending, shooting talent, and special teams are so volatile (i.e. Smith, Mason, Price, Hiller, and Anderson among others have had both amazing and terrible seasons, Clarkson's shooting percentage has fluctuated between 5 and 15% the last few seasons, and few teams sustain a good PP or PK from season to season), it's best to just leave them out of the statistic itself and throw them into the catch-all category of "luck". It's then up to the person reading the stats to identify which PDO outliers are actually "luck" and which ones might be somewhat sustainable.

The expectation is that a human looking at the numbers will say "the Avalanche have a poor Corsi and a high PDO, but Varlamov just got a goalie coach this season so he might actually have become an elite goalie, and that PDO isn't so lucky after all". Or "Stamkos has a really high PDO, but he's just a great, talented shooter and he can sustainably produce higher-quality scoring chances than the average player". Most of the people that developed these advanced stats know their limitations. The problem is that they don't do a very good job of educating the general public about their stats, so you get articles on Puck Daddy that don't seem to understand Corsi and PDO very well and draw conclusions that are beyond what the data suggest.

Thanks for the honest and helpful response. I am absolutely someone not educated enough on the stats to extrapolate anything meaningful. It has since been explained to me on the Avs subforum that the extraskater on-ice corsi stat that you used is much better than iCorsi. The list that I was linked by them and by you do indeed seem to show a much more expected list of players. I also understand that player by player is a little difficult to take seriously and you're supposed to look at it as a group but something as egregious as that was just too much to overlook. However, the on-ice Corsi numbers look a *lot* better and closer to expectations than the iCorsi numbers.

However, even with all the obvious volatility of goaltender wouldn't it be quite easy to add a "goalie effect" like they add "score effects"? Add some kind of coefficient that factors in each goalie's save% for the year into the corsi/fenwick number? It doesn't even have to be a huge modifier since the difference between an average goalie and Varly over the entire season for the Avs would have been less than 1/3 of a goal per game. Hell, use their career sv% if you want to avoid including lucky seasons or something. But I feel like (once again, uneducated knee jerk opinion so take it as is ) there could be more accurate ways than simply throwing it out because performances are volatile and often not repeatable.

 

[CanadianHockey]

Part of the problem is that people too often emphasize the predictive role of advanced stats without considering their explanatory role. They look at outliers and say 'the stats predict they will fall back to earth,' and then stop debating or thinking. Should be more critical; look at the outlier and ask, 'why are they exceeding expectations?'