Corsi Numbers

[CarltonTheBear]

Is it fair to place more value in unblocked shot attempts close vs. SAT close?

The analytics community has largely done away with "close" statistics now actually. They found that it's no better, if not worse, than just the raw numbers. Largely because it uses too small of a sample size and ignores a lot of data. It's been replaced with "score adjusted" measures that use that data in adjusted ways instead of ignoring it completely.

 

[herman]

Is it fair to place more value in unblocked shot attempts close vs. SAT close?

I don't really know.

Raw shot-attempts (Corsi) really measures what the team is doing in terms of what it can directly control. Unblocked shot attempts (Fenwick), while likely having a higher correlation to wins (?), would have the variables of shot-blocking prowess and shot-threading ability to parse out. Until it can be distilled to a clearer picture, I don't see a problem with valuing and using both.

Perhaps there is a Corsi-Fenwick differential measure that can pick out players who can get the puck on net better than most (a la Franson), or a player who can block way more than the league shot attempts against.

(Not sure about the stat labels they're using these days)
e.g. A high (Corsi60For - Fenwick60For) would indicate a shooter who could make/thread shooting lanes; a high (Corsi60Against - Fenwick60Against) would indicate a very good shot blocker.

 

[LuncheonMeat]

I think the risk with advanced statics/analytics is people looking at individual stats in a vaccuum.  It's useful information, but sometimes people point to one individual stat and say this player is great/sucks based on this one particular piece of information.  Luckily, Leafs' management has the ability to look at the big picture with players, and assess them based on many different factors, rather than ONLY advanced statistics (it seems some think this is the only thing management is basing their decisions).

I would point to what Dubas said about Fred Gauthier, as it seems like a well-rounded view on his abilities, skill and potential:

?He?s an interesting paradox,? Dubas said. ?A lot of people who are big on statistics and analytics say he hasn?t produced as much as some of these other guys . . . but if you look down to shot attempts and possession parts of it, he?s a guy that charts out extremely well.?
Dubas suggested the stats and analytics folks might have concluded the Leafs drafted him higher, 21st overall, than he deserved. But he added he believed Gauthier had a bright future.

?Mike (Babcock) uses the phrase that some guys are scout players and some guys are coach players,? Dubas said. ?Coach wants a player like Frederik on his team . . . The people who are standing on the bench, they get excited about him because he?s a guy you can trust to put out there in any circumstance. He?s great on faceoffs. He?s a big body. He plays extremely well defensively. Great stick. He?s going to be able to get more assertive defensively and get the puck out quicker.

 

[L K]

I think the risk with advanced statics/analytics is people looking at individual stats in a vaccuum.  It's useful information, but sometimes people point to one individual stat and say this player is great/sucks based on this one particular piece of information. 

It disappoints me that this even needs to be said.  Noone goes out and says "hey, stop commenting on goal scoring because it's not the only feature of a player".  For some reason analytics need to not only justify their own existence but also declare that they aren't being used exclusively as an assessment tool.

 

[Kin]

I think the risk with advanced statics/analytics is people looking at individual stats in a vaccuum.  It's useful information, but sometimes people point to one individual stat and say this player is great/sucks based on this one particular piece of information. 

It disappoints me that this even needs to be said.  Noone goes out and says "hey, stop commenting on goal scoring because it's not the only feature of a player".  For some reason analytics need to not only justify their own existence but also declare that they aren't being used exclusively as an assessment tool.

To be fair, I do think that whenever a new and useful analytic tool comes along there are some who get very into it to the point that they lose a little perspective on the bigger picture and get a little carried away with their reach or scope. I agree that ideally people like that wouldn't colour the discussion too much but frequently they're people having the discussion.

Corsi though, it's gone through that phase now. I think we've got a pretty good handle on what it measures and where its flaws are.

 

[hobarth]

"Who is the Leafs best defencemen?

In terms of Corsi statistics it is almost certainly Jake Gardiner. Over the past couple of seasons, he has boosted shot attempts for the Leafs and was the best at limiting opponents? shot attempts. This is the exact opposite of Dion Phaneuf. The majority of hockey analytics these days revolves around Corsi or other possession statistics, so the majority of hockey analytics people would probably suggest that the Leafs? best defenceman has been Jake Gardiner. Corsi wise, the conclusion is clear.

With that said, there is ample evidence that Corsi statistics significantly undervalue Phaneuf, especially defensively, and probably also over value Gardiner offensively. Add in Phaneuf?s significantly tougher quality of opponent and tougher zone start usage, and I think you can make a case that Phaneuf is the Leafs? best defenceman."

http://mapleleafshotstove.com/2015/07/05/a-statistical-look-at-the-maple-leafs-defence-entering-the-rebuild/

He's discussed pretty well every aspect that needs to be considered for the d-men but he hasn't touched much on who the d-men play with. He acknowledges Schenn played with Gardiner but didn't really mention who Phanny played with, Kostka and Holzer for instance, of course Phanny is and has been TO's best d-man and he's consistently played with subpar partners.

Corsi screams get rid of Phanny but it doesn't properly delve into the many aspects of this d-man's results in relation to who he plays with, why he's on the ice. when he's on the ice, where he starts and why. I don't feel that proper judgements can be made on whacked out corsi stats. There isn't any proper stats that can supersede the results from a qualified pair of eyes but even then inborn prejudice often colors those conclusions. Corsi is so superficial that it really doesn't provide a significant upgrade to +/-.

Corsi can only tell someone that if say Gardiner is played in ideal situations against weak opposition then he can succeed but that's not saying much.

 

[CarltonTheBear]

I promised myself I wouldn't do this...

Corsi can only tell someone that if say Gardiner is played in ideal situations against weak opposition then he can succeed but that's not saying much.

So if you accept that can't you also accept that corsi can tell someone that if say Phaneuf is played in top pairing situations against strong opposition then he can't succeed?

 

[herman]

I promised myself I wouldn't do this...

Corsi can only tell someone that if say Gardiner is played in ideal situations against weak opposition then he can succeed but that's not saying much.

So if you accept that can't you also accept that corsi can tell someone that if say Phaneuf is played in top pairing situations against strong opposition then he can't succeed?

I read this as Babcock barking it out down the bench.

Also, as mentioned by others already, Corsi (and analytics in general) is additional context, not the sole measurement of a player's ability.

 

[LuncheonMeat]

I promised myself I wouldn't do this...

Corsi can only tell someone that if say Gardiner is played in ideal situations against weak opposition then he can succeed but that's not saying much.

So if you accept that can't you also accept that corsi can tell someone that if say Phaneuf is played in top pairing situations against strong opposition then he can't succeed?

I read this as Babcock barking it out down the bench.

Also, as mentioned by others already, Corsi (and analytics in general) is additional context, not the sole measurement of a player's ability.

But that hardly makes for a good diatribe. 

 

[herman]

I promised myself I wouldn't do this...

Corsi can only tell someone that if say Gardiner is played in ideal situations against weak opposition then he can succeed but that's not saying much.

So if you accept that can't you also accept that corsi can tell someone that if say Phaneuf is played in top pairing situations against strong opposition then he can't succeed?

I read this as Babcock barking it out down the bench.

Also, as mentioned by others already, Corsi (and analytics in general) is additional context, not the sole measurement of a player's ability.

But that hardly makes for a good diatribe. 

I can't help it; with Carlton's avatar, everything he writes sounds barked now.

 

[Bullfrog]

In terms of Corsi statistics it is almost certainly Jake Gardiner. Over the past couple of seasons, he has boosted shot attempts for the Leafs and was the best at limiting opponents? shot attempts. This is the exact opposite of Dion Phaneuf. The majority of hockey analytics these days revolves around Corsi or other possession statistics, so the majority of hockey analytics people would probably suggest that the Leafs? best defenceman has been Jake Gardiner. Corsi wise, the conclusion is clear.
...

I don't think you really understand the role of analytics. I can say with almost 100% confidence that no, the majority of hockey analytics people would not suggest he's the Leaf's best defenceman, at least not based on Corsi. There may be an aggregate of statistics, video analysis, review of quality-of-competition etc. that suggests he is or isn't, but Corsi probably plays but a minor role in that.

 

[Frank E]

In terms of Corsi statistics it is almost certainly Jake Gardiner. Over the past couple of seasons, he has boosted shot attempts for the Leafs and was the best at limiting opponents? shot attempts. This is the exact opposite of Dion Phaneuf. The majority of hockey analytics these days revolves around Corsi or other possession statistics, so the majority of hockey analytics people would probably suggest that the Leafs? best defenceman has been Jake Gardiner. Corsi wise, the conclusion is clear.
...

I don't think you really understand the role of analytics. I can say with almost 100% confidence that no, the majority of hockey analytics people would not suggest he's the Leaf's best defenceman, at least not based on Corsi. There may be an aggregate of statistics, video analysis, review of quality-of-competition etc. that suggests he is or isn't, but Corsi probably plays but a minor role in that.

Enhanced statistics should just add to the equation, not replace every other performance measurement.

Well, maybe replace +/-.

 

[bustaheims]

Well, maybe replace +/-.

Let's just get rid of it. It's such a useless stat. Also, I'd support getting rid of any stat with a subjective measurement - so things like hits and such that aren't measured consistently from rink to rink.

 

[Frank E]

Well, maybe replace +/-.

Let's just get rid of it. It's such a useless stat. Also, I'd support getting rid of any stat with a subjective measurement - so things like hits and such that aren't measured consistently from rink to rink.

My personal favourite is giveaway/takeaway. 

 

[pmrules]

Well, maybe replace +/-.

Let's just get rid of it. It's such a useless stat. Also, I'd support getting rid of any stat with a subjective measurement - so things like hits and such that aren't measured consistently from rink to rink.

I'm going to regret this...but here goes.

I'm likely in the minority that loves the advanced stats and everything they represent...but I still pay attention to +/-. 

Listen I understand the love for the shiny new analytical tools - They are a great tool when used properly and in the right circumstances.  They have put a brand new look at the game which is great.  I'm not arguing this.

What I can't get my head wrapped around is the absolute hate for the "useless" +/- stat.  I don't get it. I understand that its a flawed statistic - however there are other stats (both basic and advanced) which are similarly flawed.  But what I don't understand is how come the same rules that apply to the new advanced stats (don't look at them in game/look at them over time/you need to look at context and other factors that may change the number/don't use them alone to assess a player's worth and ability) - why can't all these oversight rules apply to +/-? 

Is +/- perfect?  Absolutely not.  Does it serve a purpose?  I think it does.    I mean, I'm not saying that its the best tool or anything of measuring players - it may not even be in the top 10.  But its still useful in asking questions.  Being on the ice for more GF than GA DOES matter over time (again by applying those oversight rules which are applicable to other stats) - whether you had an impact on the play or not.   

And I think that's a lot of what's being missed in the +/- hate, is that the small plays that don't show up in the scoresheet, but that helped score a goal (like taking a hit or making a hit that opened up some of your other teammates), well, it did appear on the scoresheet since  you likely got a "+" out of it. 

I understand the not enough events and sample size arguments - and the arguments that why should a player get a "-" if they had nothing to do with a GA.  I get it.  Its not perfect.  However, why are we shutting down another potential analytical opportunity entirely?  I mean stats are here to help us ask questions and also to help us get the answers.  Here's an example of where a question can be asked about our TML players using +/- as a starting point:

What drove TML's first line (roughly speaking JVR/Bozak/Kessel) to be in the -33/34's, while its second line (roughly speaking Lupul/Kadri/Panik) was in the -7 to -10 range for the year?  They  had the same goalies and same defense and likely scored less goals as a line than the first line - What caused that significant descrepancy?  Which line provided more relative value?  I think these are valid questions.  It can be answered in many ways using other stats (TOI/QoC/Defensive vs Offensive Zone starts etc.) and by using our eyes as another measuring stick. 

(And no - i'm not looking for an answer to these questions)

But that's just my opinion and likely the last I'll speak of it - flame away.  However, here's a quote from Kevin Shattenkirk when talking about Ryan Suter here :

To be able to play as much as Suter does and still be a plus player every year is incredible. (I know you advanced stats fans don?t care about plus/minus anymore but defensemen still do.)

I mean Shattenkirk still found some use in it in roughly assessing another players ability, so it can't be all "useless". 

 

[Kin]

But what I don't understand is how come the same rules that apply to the new advanced stats (don't look at them in game/look at them over time/you need to look at context and other factors that may change the number/don't use them alone to assess a player's worth and ability) - why can't all these oversight rules apply to +/-?

I'm reminded of something Bill James wrote when he was coming up with new baseball statistics. He said that if he worked out some sort of hitting metric to determine who the best hitters were of all time and he calculated all of the players and it told him that Babe Ruth or Willie Mays or Ted Williams ranked #1 then he figured he might be on to something. If he did the calculation and it said Omar Vizquel was the best hitter of all time, he knew he'd probably made some mistakes along the way.

I think what you're talking about with +/- does hold true but the difference is that when you look at +/- with all of the caveats you're talking about(sample size, context, as a part of a larger whole) than particularly meaningful patterns don't emerge and, ultimately, that is the most important part of statistics, getting meaningful patterns from the white noise of the data.

So what does +/- mean, what is it measuring, when a player who's widely regarded as being excellent defensively can have wild swings from year to year where the context remains relatively the same? Where he does terrifically in relation to his teammates one year but terribly in relation to his teammates the next?

I think that's the really the problem with it. It doesn't measure any one thing clearly and there are no patterns that emerge from it in terms of telling us anything particularly meaningful.

I mean, take your "small plays" point. If that were true then players who excelled at making those small plays would have stronger than relative +/-'s compared to others. But I've been a fan for many years and I've really never seen a pattern like that with anyone. Everyone, it seems, more or less has a +/- that largely mirrors the quality of the team they're on.

So it's not that +/- can't be used at a starting point for asking questions, it's that it's fundamentally useless in answering any.

 

[GhostofPotvin29]

+/- is much, much, much more of a reflection of team skill than individual skill.  That's not to say it can't accurately reflect a player's actual value, but a broken clock is also right twice a day.  It's just not very reliable.

The whole reason advanced stats have gained popularity is that blogs/reporters were finding correlations between the numbers and either player or team success (depending on the stat).

And on the subject of players, former player Brian Engblom, who benefited from playing with Larry Robinson, had this to say:

?We had a strong team (in Montreal), and I led the league in 1980-81 with a plus-63, followed by a plus-78 the following year. I was always the beneficiary of a team that possessed the puck and scored goals.

Playing on a good team or a bad team impacts a player?s plus/minus ratings in a big way. I don?t care if you are a Hall of Fame player.  The best players on a bad team will really get hurt.?

Or Don Sweeney:

?The plus/minus can be a very misleading statistic. You can be totally uninvolved in a play, but there?s a breakdown some place else and the puck is in your net. Then again, you could be the sole cause of a goal against your team, but everyone else on the ice with you has to share it.?

Where I like a lot of advanced stuff, say Corsi or Fenwick, is that it's measuring shots/chances by the player - it's much more focused on what the player actually does.  But that's not to say it's not influenced by the team either.

I just think +/- is useless because there are better options available to us now.  None are perfect, but there will probably never be a perfect baseball-style stat for the NHL.  +/- in my opinion only muddies the water and is outdated.

 

[Heroic Shrimp]

Just pulling a notable name and two-way player out of a hat, Henrik Zetterberg, +/- statistics since 2009-10:

2009-10  +12
2010-11  -1
2011-12  +14
2012-13  +2
2013-14  +19
2014-15  -6

Honestly, what's any of that supposed to tell me about him as a player on the whole or season by season?

 

[Kin]

Just pulling a notable name and two-way player out of a hat, Henrik Zetterberg, +/- statistics since 2009-10:

2009-10  +12
2010-11  -1
2011-12  +14
2012-13  +2
2013-14  +19
2014-15  -6

Honestly, what's any of that supposed to tell me about him as a player on the whole or season by season?

Exactly. Even if someone wanted to go point by point there and explain how the swings make sense based on Zetterberg's linemates or the strength of the Red Wings as a whole or Zetterberg's injuries/decline through age then the fact remains that the number itself tells you so little as to be largely meaningless.

Mario Lemieux's career is another good one for just how his +/- tells you virtually nothing of consequence aside from a vague sense of it being better when the team was good and worse when the team wasn't.

 

[hobarth]

+/- is much, much, much more of a reflection of team skill than individual skill.  That's not to say it can't accurately reflect a player's actual value, but a broken clock is also right twice a day.  It's just not very reliable.

The whole reason advanced stats have gained popularity is that blogs/reporters were finding correlations between the numbers and either player or team success (depending on the stat).

And on the subject of players, former player Brian Engblom, who benefited from playing with Larry Robinson, had this to say:

?We had a strong team (in Montreal), and I led the league in 1980-81 with a plus-63, followed by a plus-78 the following year. I was always the beneficiary of a team that possessed the puck and scored goals.

Playing on a good team or a bad team impacts a player?s plus/minus ratings in a big way. I don?t care if you are a Hall of Fame player.  The best players on a bad team will really get hurt.?

Or Don Sweeney:

?The plus/minus can be a very misleading statistic. You can be totally uninvolved in a play, but there?s a breakdown some place else and the puck is in your net. Then again, you could be the sole cause of a goal against your team, but everyone else on the ice with you has to share it.?

Where I like a lot of advanced stuff, say Corsi or Fenwick, is that it's measuring shots/chances by the player - it's much more focused on what the player actually does.  But that's not to say it's not influenced by the team either.

I just think +/- is useless because there are better options available to us now.  None are perfect, but there will probably never be a perfect baseball-style stat for the NHL.  +/- in my opinion only muddies the water and is outdated.

You could also with +/- show what was a player's results when a game is close, what is a player's stats when he plays against the opposition's 1st line, 2nd line, etc. which might be more relevant than the number of shot attempts in those situations. Corsi is a shiny new toy that many have embraced and expanded on while abandoning the potential of what was already there. I see with corsi the possibilities are endless but then so are +/-. I think both corsi and +/- are more about the chance a team would have of succeeding with that player on the ice but for me corsi leaves me wanting more and even very disappointed in what it does.

I can't see how the volume of shot attempts preempts the actual figures of how many shots actually hit the net, except as a means to understand that how a player shots might explain why a high percentage of shots aren't actually successful. For example, players who insist on using the slap shot probably have fewer of them reach the opposition's net because of the length of time it takes to deliver that type of shot, I'd if I were a coach probably go to that player and tell them to learn a wrist shot which requires less time to pull off. That to me is a practical value of corsi.

Again I can't conclusively say corsi or +/- is inherently worse than the other, I personally prefer +/-, but I do understand the desire to improve evaluation tools.

Just pulling a notable name and two-way player out of a hat, Henrik Zetterberg, +/- statistics since 2009-10:

2009-10  +12
2010-11  -1
2011-12  +14
2012-13  +2
2013-14  +19
2014-15  -6

Honestly, what's any of that supposed to tell me about him as a player on the whole or season by season?

What were his corsi stats during the same seasons, I do know that Zetter has traditionally had one season on, the next off at least in his traditional stats, goals and assists, so this is probably somewhat in line with that. I would think those figures probably are similarly reflected by corsi so what both would or should say is that he or his circumstances are subject to change. It's pretty obvious that +/- couldn't help you predict  the type of year he's going to have but also neither will corsi. If you had pulled out Lidstrom instead you would probably find a far more evenly successful player that would be reasonably represented by either system.

If both ratings venues lead you to the same conclusion then why do you need both or even suppose that either is superior.