Poll: What is the most useless stat in hockey?

[ftyuv]

Let's turn it on its head: what individual stat is most important? My nomination is TOI (maybe as a percentage of possible minutes, to account for injuries) -- it's the only one that wraps all of the other stats, adds in the intangibles as measured by the coach, and says "this is how important this player to the team, according to the people who would know best."

I'll disagree on this one.  Directly, it really ONLY says "this is how important this player to the team, according to the people who would know best."  To me that is the antithesis of a statistic.  And it depends c0mpletely on The System favored by the coach.

That was actually my point. :-)  Stats are all about trying to quantify something, right? And let's cut out the middle man -- what we really want to know is, "how good is this player?"  Since it's very hard in hockey to separate that from the question of "how good is this player in this system", and since a coach/GM will presumably try in general to recruit players who match their system, I think the latter question is at the crucial intersection of Useful and Answerable. And as you yourself pointed out, TOI helps with that.

If you're looking for a hockey stat that completely objectively, and orthogonally to all other factors, tells you something useful -- then my contention is that you're going to fail. If you're looking for a hockey stat that indicates how good a player is, I think TOI is often a good one.

[jtwcornell91]

Let's turn it on its head: what individual stat is most important? My nomination is TOI (maybe as a percentage of possible minutes, to account for injuries) -- it's the only one that wraps all of the other stats, adds in the intangibles as measured by the coach, and says "this is how important this player to the team, according to the people who would know best."

I'll disagree on this one.  Directly, it really ONLY says "this is how important this player to the team, according to the people who would know best."  To me that is the antithesis of a statistic.  And it depends c0mpletely on The System favored by the coach.

That was actually my point. :-)  Stats are all about trying to quantify something, right? And let's cut out the middle man -- what we really want to know is, "how good is this player?"  Since it's very hard in hockey to separate that from the question of "how good is this player in this system", and since a coach/GM will presumably try in general to recruit players who match their system, I think the latter question is at the crucial intersection of Useful and Answerable. And as you yourself pointed out, TOI helps with that.

If you're looking for a hockey stat that completely objectively, and orthogonally to all other factors, tells you something useful -- then my contention is that you're going to fail. If you're looking for a hockey stat that indicates how good a player is, I think TOI is often a good one.

But you're not cutting out the middleman: you're making the coach the literal middleman.  It's like claiming that polls are a quantitative way of evaluating teams' performance.  All you're doing is quantifying someone's opinion of their performance.

[amerks127]

Plus-minus is a useless stat because a player on a checking line may be an excellent defensive forward, but have a terrible plus-minus number because he's usually matched against the opposition's No. 1 line. I used to cover the Hartford Whalers (remember them?) for a local paper and never put much stock in that statistic. If a goalie gives up a soft goal, for example, everyone else on the ice for his team gets a minus even though it's not their fault. There are numerous examples of how this statistic is a poor way to measure a player's value to his team. Here's a similar sentiment from an Edmonton sportswriter who did a statistical analysis of various hockey stats:
Plus/minus -- When a guy gets a plus or a minus, he might not earn it. In fact, at least 30 per cent of the time, a player has nothing to do with goal being scored, but he gets assigned a false positive or a false negative. This high error rate does little for my confidence in this stat.

The basic problem here is the player is just one of five guys on the ice, and if he's consistently out on the ice with weak players, he's going to have a weak plus/minus, no matter how good he is.The same goes for a weak player out there with good players.

You put Bobby Orr with the wrong four guys, he'll be negative plus/minus. You put Zack Stortini out there with Jason Strudwick and Kyle Brodziak, he's going to have a crappy Corsi plus/minus. The same high error rate for this stat applies to all plus/minus stats, be they Shots For/Shots Against plus/minus, or Corsi plus/minus (which is Shots At For/Shots At Against plus/minus), or Scoring Chances For/Scoring Chances Against plus/minus.

The full article is available here: http://communities.canada.com/edmontonjournal/print.aspx?postid=359038

I think David put it most succinctly.  I also voted for +/- not just because I can corroborate his comments from other scouts, but because of the attention fans and analysts give to +/-.  If you look at tsn.ca's statistics page (http://tsn.ca/nhl/statistics/) you'll find +/- listed, but not game-winning goals.  I think a better question for this poll would be "What is the most overrated stat in hockey" and in that regard it has to be +/- because no really pays attention to GWG.

[ftyuv]

Let's turn it on its head: what individual stat is most important? My nomination is TOI (maybe as a percentage of possible minutes, to account for injuries) -- it's the only one that wraps all of the other stats, adds in the intangibles as measured by the coach, and says "this is how important this player to the team, according to the people who would know best."

I'll disagree on this one.  Directly, it really ONLY says "this is how important this player to the team, according to the people who would know best."  To me that is the antithesis of a statistic.  And it depends c0mpletely on The System favored by the coach.

That was actually my point. :-)  Stats are all about trying to quantify something, right? And let's cut out the middle man -- what we really want to know is, "how good is this player?"  Since it's very hard in hockey to separate that from the question of "how good is this player in this system", and since a coach/GM will presumably try in general to recruit players who match their system, I think the latter question is at the crucial intersection of Useful and Answerable. And as you yourself pointed out, TOI helps with that.

If you're looking for a hockey stat that completely objectively, and orthogonally to all other factors, tells you something useful -- then my contention is that you're going to fail. If you're looking for a hockey stat that indicates how good a player is, I think TOI is often a good one.

But you're not cutting out the middleman: you're making the coach the literal middleman.  It's like claiming that polls are a quantitative way of evaluating teams' performance.  All you're doing is quantifying someone's opinion of their performance.

The crucial difference being that polls ask people their opinions about teams they don't actually know very well, whereas coaches know their players extremely well.

You're right about the middle man bit, though. I should have instead said that since stats are a middleman anyway, let's at least pick the right middleman. The point is that we should focus on what we want the stats to tell us, not just on the stats for the sake of their objective purity.

[David Harding]

Let's turn it on its head: what individual stat is most important? My nomination is TOI (maybe as a percentage of possible minutes, to account for injuries) -- it's the only one that wraps all of the other stats, adds in the intangibles as measured by the coach, and says "this is how important this player to the team, according to the people who would know best."

I'll disagree on this one.  Directly, it really ONLY says "this is how important this player to the team, according to the people who would know best."  To me that is the antithesis of a statistic.  And it depends c0mpletely on The System favored by the coach.

That was actually my point. :-)  Stats are all about trying to quantify something, right? And let's cut out the middle man -- what we really want to know is, "how good is this player?"  Since it's very hard in hockey to separate that from the question of "how good is this player in this system", and since a coach/GM will presumably try in general to recruit players who match their system, I think the latter question is at the crucial intersection of Useful and Answerable. And as you yourself pointed out, TOI helps with that.

If you're looking for a hockey stat that completely objectively, and orthogonally to all other factors, tells you something useful -- then my contention is that you're going to fail. If you're looking for a hockey stat that indicates how good a player is, I think TOI is often a good one.

But you're not cutting out the middleman: you're making the coach the literal middleman.  It's like claiming that polls are a quantitative way of evaluating teams' performance.  All you're doing is quantifying someone's opinion of their performance.

The crucial difference being that polls ask people their opinions about teams they don't actually know very well, whereas coaches know their players extremely well.

You're right about the middle man bit, though. I should have instead said that since stats are a middleman anyway, let's at least pick the right middleman. The point is that we should focus on what we want the stats to tell us, not just on the stats for the sake of their objective purity.

Some of us would like to have stats to quantify "leadership," "hustle," "coming to play," and the all-important "heart.":-D  I think we agree long ago that you can't measure those things.  It hard to argue that TOI is not a good measure of the coaches' perception of those characteristics.  
Maybe we should do a correlation study.  Are those whose ideal is an accurate statistical description of a player more likely to be engineers and physical sciences majors?

[jtwcornell91]

Some of us would like to have stats to quantify "leadership," "hustle," "coming to play," and the all-important "heart.":-D

Hark, is that the sound of Sabermetricians sharpening their knives? ::uhoh::

[adamw]

I'm sorry I missed this conversation when it started.  I respectfully disagree with the sentiment on numerous counts.

NO statistic is useless. Every measurement can tell you something. It is not the fault of the statistic if it's misinterpreted by humans.

Someone (Trotsky) mentioned Sabermetrics / Bill James.  Bill James makes this point when he defends the "save" and the notion of "quality starts" ... He used that once as a jumping off point to a lengthy article on the usefulness of stats.  Yes, it's true, some statistics are more indicative than others.  Yes, it's true that some statistics are more susceptible to meaningless influences than others.  But it doesn't render a statistic necessarily meaningless.

Someone mentioned the Game-Winning RBI in baseball. It was eliminated because it was deemed useless. And it indeed had numerous flaws, just as the Game Winning Goal does. But neither is completely useless. ... In fact, baseball's and hockey's were/are determined differently. Each is consistent within its sport, and each has its plusses and minuses.  I remember when GWRBI was being panned -- people said it should be more like hockey. In hockey, the GWG is the goal that's +1 the other team's total. In baseball, the GWRBI was the RBI that put your team ahead to stay. This is consistent within each sport, because that's the way a goaltender's win is calculated vs. a pitcher's win.  But it's not accurate to say one is inherently a better method than the other -- and it used to drive me crazy to hear people criticize the GWRBI when they weren't criticizing the way a pitcher's win was calculated.  In either method, there are going to be flawed Game Winners, and "good" Game Winners.  Neither is perfect, but it doesn't have to be.

James' argument was that no statistic is perfect in a team sport -- they are all subject to flaws. How many goals are empty netters?  How many go off a guy's skate and in?  Obviously, the goal statistic is "less flawed," as a whole, because even if they go off a skate, at least you can say it directly contributed to your team winning. As do RBIs in baseball. However, RBIs are still dependent on your team having runners on base. Martin St. Louis' and Eric Perrin's linemate at Vermont was J.C. Ruid -- and he scored a lot of goals thanks to them. Would he have scored those goals playing on the 4th line?  Hell no.

Next point ... I actually do believe there are clutch hockey players, and clutch baseball players.  I believe this despite being an ardent supporter of Bill James/Sabermetrics.  I have seen players at the top of Game Winning goal statistics consistently each year -- and not just because they score a lot in general.  Problem is, you're dealing with small sample sizes -- so it's hard to tell statistically. But to say there aren't players that rise to an occasion, is really quite silly to anyone who actually watches sports.

I don't believe in "momentum" - but I do believe in "clutch" ... I think they are two different things.

Finally ... the plus-minus statistic. I am an ardent supporter of this statistic for being very useful. Yes, of course, you might have just lolly-gagged on the ice, and get credit with a plus just because you were there when Ovechkin went nuts. But so what. Again, any stat is open to this kind of thing. Assists are dependent upon the other guy scoring the goal. Runs are dependent on someone driving you in. Wins and shutouts are dependent upon your defense. Etc....

We have to assume when looking at something like plus-minus that -- despite the fact that you get some cheapies, that EVERYONE gets cheapies. And therefore, over the course of the season, that plays itself out. Similarly, obviously a player on a bad team will have a generally lower plus-minus total. That is why us, as interpreters, must look at the stat and take it within the context of the team. Why is this so easy for people to do when it comes to Wins for a pitcher?  They know that a great pitcher on a crappy team wins 15 games, and it means he's awesome. They are able to adjust that in their minds.  Same thing for plus-minus.  No one is saying that the 3rd liner on the Penguins is better than the star on Edmonton just because the Penguins guy has a better plus-minus. But you can make that adjustment relatively, and come up with a good indication.

I am very much a supporter that plus-minus is a good indicator of someone's general two-way worth to a team. And the best players on the best teams have the best plus-minus.

So I take umbrage with the idea than any statistic is useless. It's all in how we choose to apply and interpret those statistics. Useful information comes from all of it.

Peace out.

[ugarte]

I actually do believe there are clutch hockey players, and clutch baseball players.  I believe this despite being an ardent supporter of Bill James/Sabermetrics.  I have seen players at the top of Game Winning goal statistics consistently each year -- and not just because they score a lot in general.  Problem is, you're dealing with small sample sizes -- so it's hard to tell statistically. But to say there aren't players that rise to an occasion, is really quite silly to anyone who actually watches sports.

To the first point, I'd like to see a citation about the players that were consistently at the top of the GWG stats who weren't otherwise top scorers. The entirety of the opposition to the notion of clutch comes from noticing that there is almost never any consistency to those kind of stats from year to year.

To the second point, that people who actually watch sports wouldn't agree that there are players who rise to the occasion, I'd say that players get a reputation and once they do, anything that follows that matches up "proves" the initial impression and anything that doesn't match up is written off and quickly forgotten. Evgeny Malkin was a choker in 2008 but a pressure performer in 2009. Sportswriters are scratching their heads trying to figure out how Tony Romo and Alex Rodriguez suddenly stopped being chokers just as they once did with Barry Bonds. Donovan McNabb, who has been to a Super Bowl, will never shake his reputation unless he gets himself a ring. In the end it is 90% bullshit and 10% post-hoc mythmaking.

[Jim Hyla]

NO statistic is useless. ...

Someone mentioned the Game-Winning RBI in baseball. It was eliminated because it was deemed useless. And it indeed had numerous flaws, just as the Game Winning Goal does. But neither is completely useless. ... ::wtf::
 
In hockey, the GWG is the goal that's +1 the other team's total.

Well, the thread was titled "What is the most useless stat in hockey?", not completely useless; but this one has to come close. We're up 3-1 and I score. We give up 2 goals and I'm a winner?::screwy::

Peace out.

With this I can agree.

[Ken70]

KRACH

[adamw]

To the second point, that people who actually watch sports wouldn't agree that there are players who rise to the occasion, I'd say that players get a reputation and once they do, anything that follows that matches up "proves" the initial impression and anything that doesn't match up is written off and quickly forgotten. Evgeny Malkin was a choker in 2008 but a pressure performer in 2009. Sportswriters are scratching their heads trying to figure out how Tony Romo and Alex Rodriguez suddenly stopped being chokers just as they once did with Barry Bonds. Donovan McNabb, who has been to a Super Bowl, will never shake his reputation unless he gets himself a ring. In the end it is 90% bullshit and 10% post-hoc mythmaking.

I would generally agree with you .... I just believe there are people who can rise to the occasion. Just from playing sports myself, I have seen players that wilt under pressure, and seen players who do better under it.  Does it play itself out enough in statistics to be noticed -- that's the question.  And I know Bill James et al says that it can all be explained through random fluctuation.  I get it.  I agree....

Anyway .... Jim's right .... the discussion was about "most useless" not totally useless.  So OK.  On that note, you win

[ftyuv]

To the second point, that people who actually watch sports wouldn't agree that there are players who rise to the occasion, I'd say that players get a reputation and once they do, anything that follows that matches up "proves" the initial impression and anything that doesn't match up is written off and quickly forgotten. Evgeny Malkin was a choker in 2008 but a pressure performer in 2009. Sportswriters are scratching their heads trying to figure out how Tony Romo and Alex Rodriguez suddenly stopped being chokers just as they once did with Barry Bonds. Donovan McNabb, who has been to a Super Bowl, will never shake his reputation unless he gets himself a ring. In the end it is 90% bullshit and 10% post-hoc mythmaking.

I would generally agree with you .... I just believe there are people who can rise to the occasion. Just from playing sports myself, I have seen players that wilt under pressure, and seen players who do better under it.  Does it play itself out enough in statistics to be noticed -- that's the question.  And I know Bill James et al says that it can all be explained through random fluctuation.  I get it.  I agree....

Anyway .... Jim's right .... the discussion was about "most useless" not totally useless.  So OK.  On that note, you win

The problem I'm hearing with GWGs (and it's one I agree with) isn't that there's no such thing as clutch players.  It's that the goal that ends up being the GWG is often not a clutch one. If a team is up 2-0, the goal that puts them 3-0 isn't exactly high pressure; but if the other team scores two late in the third, suddenly that third goal is the clutch, high pressure GWG. I agree with Robb that the go-ahead or tying goal is much more interesting, even if it's diluted by virtue of the fact that there can be several of them in a game.

[Tom Lento]

I would generally agree with you .... I just believe there are people who can rise to the occasion. Just from playing sports myself, I have seen players that wilt under pressure, and seen players who do better under it.  Does it play itself out enough in statistics to be noticed -- that's the question.  And I know Bill James et al says that it can all be explained through random fluctuation.  I get it.  I agree....

Anyway .... Jim's right .... the discussion was about "most useless" not totally useless.  So OK.  On that note, you win

I'm way too lazy to look it up, but someone actually did an analysis of "clutch" players and determined that, with one exception, none of the clutch performers were actually clutch at all. Statistically speaking, the clutch performances for players was within their normal range. I don't think anyone has done an "anti-clutch" analysis. Personally, I find it far easier to believe that some players consistently choke under pressure than that some players are basically mediocre and then consistently do heroic things when the pressure is high. I don't believe in clutch, and I see no evidence to support it, but I haven't seen any evidence about choking and it makes intuitive sense that it would happen.

[Jim Hyla]

I would generally agree with you .... I just believe there are people who can rise to the occasion. Just from playing sports myself, I have seen players that wilt under pressure, and seen players who do better under it.  Does it play itself out enough in statistics to be noticed -- that's the question.  And I know Bill James et al says that it can all be explained through random fluctuation.  I get it.  I agree....

Anyway .... Jim's right .... the discussion was about "most useless" not totally useless.  So OK.  On that note, you win

I'm way too lazy to look it up, but someone actually did an analysis of "clutch" players and determined that, with one exception, none of the clutch performers were actually clutch at all. Statistically speaking, the clutch performances for players was within their normal range. I don't think anyone has done an "anti-clutch" analysis. Personally, I find it far easier to believe that some players consistently choke under pressure than that some players are basically mediocre and then consistently do heroic things when the pressure is high. I don't believe in clutch, and I see no evidence to support it, but I haven't seen any evidence about choking and it makes intuitive sense that it would happen.

So, there is no Mr. October?:-}

[RichH]

I would generally agree with you .... I just believe there are people who can rise to the occasion. Just from playing sports myself, I have seen players that wilt under pressure, and seen players who do better under it.  Does it play itself out enough in statistics to be noticed -- that's the question.  And I know Bill James et al says that it can all be explained through random fluctuation.  I get it.  I agree....

Anyway .... Jim's right .... the discussion was about "most useless" not totally useless.  So OK.  On that note, you win

I'm way too lazy to look it up, but someone actually did an analysis of "clutch" players and determined that, with one exception, none of the clutch performers were actually clutch at all. Statistically speaking, the clutch performances for players was within their normal range. I don't think anyone has done an "anti-clutch" analysis. Personally, I find it far easier to believe that some players consistently choke under pressure than that some players are basically mediocre and then consistently do heroic things when the pressure is high. I don't believe in clutch, and I see no evidence to support it, but I haven't seen any evidence about choking and it makes intuitive sense that it would happen.

So, there is no Mr. October?:-}

In fact, doesn't that prove the point?  Reggie Jackson is a HOF'er regardless of the month.  That he excelled in October is no surprise because he was an excellent power hitter normally.

I think the one thing I believe in that staunch Sabermetricians don't is "streakiness." Timing, muscle memory, and physical well-being can come in and out of sync, I've felt that feeling where the ball seems like a beach ball, or I know I'm hitting the sweet-spot every time.  I believe in "streakiness" more than I do "clutch."