Why I don't like shots on goal as indicator of individual performa

[Daniel]

It's not a secret to anyone.  I dislike shots on goal as a statistic, especially when it comes to individual performance.  Note, I don't think it's totally irrelevant, just overstated.  My hypothesis is that shots on goal will pretty closely reflect time on ice across all players, from the good to the not so good (goons like Barch do not get included).  So I decided to do an off the cuff analysis of shots on goal per minute of ice time for certain Devils for this season.  I picked Clarkson, Gionta, Carter, Kovalchuk, Zajac, Elias and Henrique.  I think it's a decent distribution of different types of players for purposes of a message board post.

Here are the numbers, rounded to the nearest hundredth of a shot per minute on ice. 

Clarkson: .21

Gionta:  .09

Carter: .11

Kovalchuk:.13

Zajac: .09

Elias:  .13

Henrique:  .1

You will notice that the difference between Gionta, Carter, Kovalchuk, Zajac, Elias, Henrique are noticeable, but there isn't a huge divergence.  Clarkson is the outlier, but of course, there's an explanation.  He plays on the powerplay, but not on the penalty kill.  Kovalchuk, Elias and Zajac get significant powerplay and penalty kill time.  Gionta and Carter, honestly, I don't know for sure.  I looked around quickly for even strength/pk/pp icetime per player, but could not find anything, although I believe Gionta gets significant PK time, but neither a lot of PP time. 

What this tells me is that shots on goal are similar to points per game in basketball.  Any basketball team, whether it's the Heat or the Bobcats,
are going to have at least one or two players that will put up respectable points per game.  Basically, someone has to score the points, same as someone has to get the shots in hockey.  In hockey, how many shots you’re getting relative to everyone else is more a product of icetime and type of icetime than anything else.

Now, we hear a lot about the shot differential for the Devils as a team this year, a lot more shots for as opposed to shots against.  Just a guess on my part, and I don't know if it can be measured, but the Devils style of play is designed to put as many shots on goal as possible with less regard for seeking out high percentage shots.  This is especially the case on the powerplay this year, as opposed to the Oates powerplay.  Ultimately, the way I see it, is that not all shots by all teams and players are created equal when it comes to the "purpose" of a shot (just taking a shot for its own sake with less regard for scoring opportunity) or situation in which a shot is taken.  This is why I think shooting percentage is overrated as well.  Ultimately, you have to watch the games.

Conclusion,I still think the best measure of a forward is how many goals a forward creates, either goals scored or "meaningful" assists.  I know some people try to calculate quality scoring chances, but it's too imprecise to mean anything so far as I'm concerned.

[cgb6397]

To me, it doesn't get much better than corsi. However, hockey is a sport in which I try not to read too much into numbers. The best measure of a player, are goals created,corsi,hits and of course, watching their game.

[Derlique]

Interesting, but these numbers mean nothing if there's no context behind them. For example, Elias' numbers are wayyy more impressive than Gionta's because PE plays against top lines while Gionta plays against 4th lines and 3rd pairs.

[Daniel]

Interesting, but these numbers mean nothing if there's no context behind them. For example, Elias' numbers are wayyy more impressive than Gionta's because PE plays against top lines while Gionta plays against 4th lines and 3rd pairs.

But again, Elias plays in the PP. Also, CBGB plays the role of checking line, and will go up against the better players in other teams, how often viz Elias, I can't tell you. I seem to recall a long discussion about Gionta unjustifiably getting "tough" minutes.

[capo]

I think individual shots represent a player far better than team shots represent a team.   A player's quantitiy tells about their individual game and how involved they were.  Where team shots are concerned I think quality is of far greater importants.  I like greedy goals scorers.  A lot of times a player will pass up a good shot to look unselfish.  Jason Spezza is a great example of this.

[sundstrom]

here's the issue - if you give stephen gionta 23 minutes a game, he's not going to keep that shots/toi up because he's going to get unfavorable matchups where he's never in a position to score.

 

shots/60 does a better job of qualifying who is a good player and who isn't if you want to work shots into the equation and it shows a much different picture that what you paint - which it should because nobody is arguing that gionta is in the same neighborhood as zajac in terms of offensive ability.

[Daniel]

here's the issue - if you give stephen gionta 23 minutes a game, he's not going to keep that shots/toi up because he's going to get unfavorable matchups where he's never in a position to score.

shots/60 does a better job of qualifying who is a good player and who isn't if you want to work shots into the equation and it shows a much different picture that what you paint - which it should because nobody is arguing that gionta is in the same neighborhood as zajac in terms of offensive ability.

I'm trying to remember my algebra, but shots per 60 will get you the same result as shots per time on ice across players.

And yes, my point is that we all know Gionta is not in the same neighborhood as Zajac. You would know that by watching games and looking at goals produced, rather than shots.

[Devils731]

We are using a pretty small sample size and only 7 players, so I don't think this can really tell us anything about any general trend.

 

I think this exercise is a little more precise if we look at even strength SOG/60, rather than overall shots per minute.  The numbers are just bigger and easier as SOG/60 and limiting to even strength is good idea.

 

I added a few names in as well, but it certainly doesn't make the examination much more broad or telling.

 

2012-2013

 

Clarkson: 11.88
Gionta: 5.11
Carter: 6.58
Kovalchuk: 6.27
Zajac: 5.23
Elias:  7.93
Henrique:  5.90

Sestito:  5.09

Zubrus:  3.65

Poni:  5.82

 

2011-2012

 

Clarkson: 10.33
Gionta:  N/A
Carter:  4.36
Kovalchuk:  8.61
Zajac:  5.71
Elias:   6.29
Henrique:  5.38

Sestito:  2.48

Zubrus:  4.33

Poni:  7.61

 

--------------------------------

 

So nothing really definitive on this list, but it does appear that players who score more are also the players who tend to generate more shots on goal.

 

Also, I would say most people don't like SOG as a shorthand way to look at things anyway.  Most number-y people will want to look at shots on goals + missed shots or shots on goal + missed shots + blocked shots to determine which players had good or bad seasons, when they couldn't watch a player or team very much.

[SterioDesign]

I think Triumph and I should not take part in this discussion or we'll get in yet another hardcore debate

[Triumph]

Goals over the short run are notoriously unpredictable.  And by the short run, I basically mean 'at least 2 years'.  Over a career you've got the problem of changing shooting percentages - they rise and fall with age.  So too do shot rates - Oiler fans on the microstats train were shouting up and down that Jordan Eberle couldn't maintain his goals or points production.  He didn't, but he improved as a player and got more shots on goal, and maybe if he improves some more he'll be as 'good' as the player who shot nearly 19% over a full season.

 

This is one of those issues that seems to plague certain people, but goals are obviously the 'value' - you don't win games with shots on goal, you win it with goals.  The trouble is, shots on goal are often more predictive of who's going to be a goal scorer in future than goals themselves, as shooting percentage tends to regress towards a personal mean, but shot rates are more stable than shooting percentages.  If you're evaluating players for free agency and trade, there's no need to consider their past value, only their future value.

Edited May 21, 2013 by Triumph

[Daniel]

Goals over the short run are notoriously unpredictable.  And by the short run, I basically mean 'at least 2 years'.  Over a career you've got the problem of changing shooting percentages - they rise and fall with age.  So too do shot rates - Oiler fans on the microstats train were shouting up and down that Jordan Eberle couldn't maintain his goals or points production.  He didn't, but he improved as a player and got more shots on goal, and maybe if he improves some more he'll be as 'good' as the player who shot nearly 19% over a full season.

 

This is one of those issues that seems to plague certain people, but goals are obviously the 'value' - you don't win games with shots on goal, you win it with goals.  The trouble is, shots on goal are often more predictive of who's going to be a goal scorer in future than goals themselves, as shooting percentage tends to regress towards a personal mean, but shot rates are more stable than shooting percentages.  If you're evaluating players for free agency and trade, there's no need to consider their past value, only their future value.

 

I guess I'm a bit lost.  If you're a GM evaluating Clarkson for UFA, are you really going to predict that he'll score anymore than 15 to 20 goals a year.  He gets a ton of shots, but in the end, he's still David Clarkson -- a gritty, but not particularly skilled player.  Or does the answer lie in the fact that he's getting older, and won't be as good a shooter?  Or does this account for your qualifier that shots are "often" more predictive. 

 

Also, I would say that assists get overlooked as well when people focus too much on shots.   

[Devils731]

Clarkson scored 30 and 25(pro-rated) goals the last 2 years. If goals are the best measure of a player, why would you then peg him as a 15-20 goal guy, when that's well below what's he's done his like 130 games.

No stat in something as random as hockey is going to predict the future perfectly. No one number perfectly sums up a hockey player. You need to use multiple numbers to support why you believe something about a player or a team.

When your eyes and the numbers agree on something then you've probably found a nugget of truth. When your eyes and numbers don't agree then one of those is wrong and you need to decide which.

Daniel, I guess I'm struggling with what you're trying to say or explore.

[StormJosh]

I think you HAVE to use ES SOG/60 (or ES CORSI) to accurately compare forwards.

 

I think the main argument in favor of using a statistic like Shots on Goal per 60 minutes of Even-Strength ice time combined with Zone-Start percentage, Quality of Competition, and Quality of Linemates allows you to make a MUCH better evaluation of a player rather than goals.

 

Goals incorporates too many things that are out of the players control. For example, the player could be generating a ton of high quality scoring chances and running into incredible goaltending. Or, the player may have terrible linemates. Or the player could be named Kesler and start and be facing the oppositions best players all the time from the defensive zone which reduces a players effectiveness if we only evaluate goals.

 

Now, if we look at the previous stats, Even-Strength Shots on Goals per 60 minutes of ice time eliminates bias towards PP/SH specialists and ice time. Next, Zone-Start percentage tells us if we should "adjust" our perception of this player when looking at SOG/60 favorably or unfavorably depending on zone-start. If a player is always starting in the defensive zone, it is naturally more difficult to generate offense than starting in the attacking zone. Finally, quality of competition and quality of linemates can again provide an adjustment for obvious reasons.

 

Lets compare two players.

 

Player A:

10 G, 22 A, 32 P

ES Corsi: 26.62 (higher is better)

Off-Zone %: 42.4 (higher means more offensive zone starts)

QoC: 0.627 (higher is more challenging)

 

Player B:

13G, 36A, 49 P

ES Corsi: -10.6

Off-Zone %:50.8

QoC: -0.3

 

By goals and assists, Player B is far superior. By all other measures, Player A is destroying Player B.

 

Player A is Patrice Bergeron

Player B is Mike Ribiero

 

Who is the player you would rather have?

 

EDIT: All numbers above, except total goals and assists are ES numbers.

Edited May 21, 2013 by StormJosh

[Daniel]

Clarkson scored 30 and 25(pro-rated) goals the last 2 years. If goals are the best measure of a player, why would you then peg him as a 15-20 goal guy, when that's well below what's he's done his like 130 games.

No stat in something as random as hockey is going to predict the future perfectly. No one number perfectly sums up a hockey player. You need to use multiple numbers to support why you believe something about a player or a team.

When your eyes and the numbers agree on something then you've probably found a nugget of truth. When your eyes and numbers don't agree then one of those is wrong and you need to decide which.

Daniel, I guess I'm struggling with what you're trying to say or explore.

 

Basically, my hypothesis is that shots on goal tends to have a close correlation to time on ice, across virtually all types of players.  So when someone says X is good because he gets a lot of shots, you're really just saying that X gets more ice time (or relatively more/less PP/PK time) than Y, which doesn't tell you all that much.  If you could theoretically flip the icetime that a guy like Kovalchuk gets versus someone like Carter, Carter would get more shots, but Kovalchuk will likely still have more goals, which would tell you that Kovalchuk is a better player if you didn't watch the games.

[Devils731]

Basically, my hypothesis is that shots on goal tends to have a close correlation to time on ice, across virtually all types of players. 

 

I don't think this is true.  Players who score more also tend to get more shots with their ice time.

 

If you could theoretically flip the icetime that a guy like Kovalchuk gets versus someone like Carter, Carter would get more shots, but Kovalchuk will likely still have more goals, which would tell you that Kovalchuk is a better player if you didn't watch the games.

 

Kovaluchuk, for his career, is an oddball.  He's one of the very few handful of players who has been able to shoot a well above league average shooting percentage.  The vast majority of forwards to tend fall into the same range as each other, within a couple of percentage points.  So Kovy always looks weird when comparing players because usually shooting percentage is ignored since most players end up around with around the same number.

 

Last season Kovalchuk generated almost twice as many shots as Carter per minute.  So no, flipping Carter and Kovalchuk is likely to leave your team with many less overall shots on goal and many less goals total.

 

--------------------------------------------------------------

 

http://blogs.thescore.com/nhl/2012/03/24/the-nhls-leaders-in-shots-per-60-minutes-of-play/

 

That article has a nice chart on the players who were the best at generating SOG per 60 through most of last season.  It's not a perfect list of amazing players, but you'll see you don't end up with crappy players on the list, which we should see if SOG is solely a function of ice time.  

 

I think this list and the Devils numbers from last season show that this short season for the Devils was most likely the outlier and not the norm.

[Daniel]

I don't think this is true.  Players who score more also tend to get more shots with their ice time.

 

 

Kovaluchuk, for his career, is an oddball.  He's one of the very few handful of players who has been able to shoot a well above league average shooting percentage.  The vast majority of forwards to tend fall into the same range as each other, within a couple of percentage points.  So Kovy always looks weird when comparing players because usually shooting percentage is ignored since most players end up around with around the same number.

 

Last season Kovalchuk generated almost twice as many shots as Carter per minute.  So no, flipping Carter and Kovalchuk is likely to leave your team with many less overall shots on goal and many less goals total.

 

--------------------------------------------------------------

 

http://blogs.thescore.com/nhl/2012/03/24/the-nhls-leaders-in-shots-per-60-minutes-of-play/

 

That article has a nice chart on the players who were the best at generating SOG per 60 through most of last season.  It's not a perfect list of amazing players, but you'll see you don't end up with crappy players on the list, which we should see if SOG is solely a function of ice time.  

 

I think this list and the Devils numbers from last season show that this short season for the Devils was most likely the outlier and not the norm.

 

Do you have anything that controls for powerplay/penalty kill time? 

[Devils731]

Do you have anything that controls for powerplay/penalty kill time? 

 

 

All the numbers I used were for ES only.

 

http://www.behindthenet.ca/nhl_shot_statistics.php?ds=13&s=13&f1=2012_s&f2=5v5&f5=N.J&c=0+1+3+5+7+9+10+11+12+13

 

That page will get you most of the raw numbers I used if you want to play around with them or look at other seasons.

[Triumph]

Carter and Kovalchuk's shots/60 were equal this year at 5 on 5, 731, so I guess you must be referring to 2011-12?

[Devils731]

Carter and Kovalchuk's shots/60 were equal this year at 5 on 5, 731, so I guess you must be referring to 2011-12?

 

"Last season Kovalchuk generated almost twice as many shots as Carter per minute."

 

I still consider this season the 2012-2013 season, so last season, whenever I mentioned it, was 2011-2012.

[EdgeControl]

Kovy either misses the net, hits the post or scores LOL   that leads to a high %