Nov 222011
 

I hate to keep beating the “Shooting Percentage Matters” drum but it really dumbfounds me why so many people choose to ignore it, or believe it is only a small part of the game and not worth considering and instead focus their attention on corsi/fenwick, and corsi/fenwick derived stats as their primary evaluation too.

It dumbfounds me that people don’t think players have an ability to control shooting percentage yet we all seem to agree that shooting percentage is affected by game score.  Rob Vollman wrote the following in a comment thread at arctic ice hockey.

<blockqote>The score can affect the stats because teams behave differently when chasing or protecting a lead…</blockquote)

He isn’t specifically referring to shooting percentage, but shooting percentage varies based on game score and I think most people accept that.  So, while people freely accept that teams can play differently depending on score, they seemingly choose not to believe that players can play different depending on their role, or skillset.  Or rather, it isn’t that they don’t believe players can play differently (for example they realize there are defensive specialists) they just choose not to accept that a players style of play (in addition to their talents, which often dictates their style of play) will affect their stats, including shooting percentage.  An example, which I brought up at The Puck Stops Here is Marian Gaborik vs Chris Drury.  Both Gaborik and Drury played the past 2 seasons on the NY Rangers but Gaborik played an offensive role and Drury generally played a more defensive/3rd line role.  As a result, here are their offensive stats at 5v5 over the past 2 seasons.

Gaborik Drury Gaborik’s Edge
Team Fenwick For per 20min WOI 13.8 12.8 +8%
Team Sh% For WOI 10.26% 6.18% +66%
Team Goals For per 20 min WOI 1.031 .575 +79%

Shooting percentage took what was a slight edge for Gaborik in terms of offensive fenwick for and turned it into a huge advantage in goals for.  Part of that is Gaborik and his line mates better skill level and part of it is their aggressive offensive style of play, but regardless of why, we need to take shooting percentage into account or else we will undervalue Gaborik at the offensive end of the rink and over value Drury.

It isn’t just Gaborik and Drury whose offense is significantly impacted by shooting percentage.  It happens all the time.  I took a look at all players that had 2000 5v5 even strength on-ice offensive fenwick events over the past 4 seasons.  From there I calculated their expected on-ice goals scored based on their ice time using league-wide average  on-ice fenwick for per 20 minutes (FF20) and league-wide average fenwick shooting percentage (FSH%).

I next calculated an expected goals based on the league-wide FF20 and the players FSH% as well as an expected goals based on the players FF20 and the league-wide average FSH%.  When we compare these expected goals to the expected goals based solely on the league-wide average we can get an idea of whether a players on-ice goal production is driven mostly by FF20 or FSH% or some combination of the two.

The following players had their on-ice 5v5 goal production influenced the most positively or most negatively due to their on-ice 5v5 FSH%.

Player Name %Increase from FSH%
MARIAN GABORIK 40.6%
SIDNEY CROSBY 36.3%
ALEX TANGUAY 33.1%
HENRIK SEDIN 32.8%
BOBBY RYAN 32.5%
EVGENI MALKIN 31.9%
DANIEL SEDIN 31.6%
ILYA KOVALCHUK 30.6%
NATHAN HORTON 29.6%
J.P. DUMONT 29.4%
GREGORY CAMPBELL -12.4%
RYAN CALLAHAN -13.9%
RADEK DVORAK -15.6%
CHRIS DRURY -16.8%
SEAN BERGENHEIM -19.4%
SCOTT GOMEZ -19.7%
MARTIN HANZAL -21.5%
MIKE GRIER -21.5%
DANIEL WINNIK -24.5%
TRAVIS MOEN -32.1%

And the following players had their on-ice 5v5 goal production influenced the most positively or most negatively due to their on-ice 5v5 FF20.

Player Name %Increase from FF20
HENRIK ZETTERBERG 24.7%
ALEX OVECHKIN 21.7%
PAVEL DATSYUK 20.6%
TOMAS HOLMSTROM 19.9%
NICKLAS BACKSTROM 19.8%
ERIC STAAL 19.7%
RYANE CLOWE 18.8%
ALEXANDER SEMIN 18.3%
SCOTT GOMEZ 18.0%
ZACH PARISE 17.9%
MARTY REASONER -6.5%
ANDREW COGLIANO -6.5%
ANTTI MIETTINEN -6.7%
KYLE BRODZIAK -7.3%
CHRIS KELLY -8.6%
ILYA KOVALCHUK -9.8%
JAY MCCLEMENT -10.4%
MICHAL HANDZUS -14.4%
JOHN MADDEN -14.5%
TRAVIS MOEN -15.6%

Some interesting notes:

  1.  The range in the influence of FSH% is significantly larger than the range of influence of FF20 indicating that shooting percentage is more important than shot generation in terms of scoring goals.
  2. The FSH% list is not random.  The list is stratified.  Offensive players at the top, non-offensive players at the bottom (plus Scott Gomez who gets offensive minutes, but sucks).  What you see above is not luck.  There is order to the list, not randomness.
  3. Speaking of Gomez, he sucks at on-ice FSH%, but has a very good FF20, though that is partly due to offensive zone start bias.
  4. Ilya Kovalchuk is the anti-Gomez.  He has a great FSH%, but is horrible at helping his team generate shots.
  5. The standard deviation of the FSH% influence is 14.5% while it is 8.3% for FF20 influence so it seems FSH% has a much greater influence on scoring goals than FF20.  This is not inconsistent with some of my observations in the past or observations of others.

So, what does all this mean?  Shooting percentage matters, and matters a lot and thus drawing conclusions based solely on a corsi analysis is flawed.  It isn’t that generating shots and opportunities isn’t important, but that being great at it doesn’t mean you are a great player (Gomez) and being bad at it doesn’t make you a bad player (Kovalchuk).  For this reason I really cringe when I see people making conclusions about players based on a corsi analysis.  A corsi analysis will only tell you how good he is at one aspect of the game, but is not very good at telling you the players overall value to his team.  My goal is, and always will be, to try and evaluate a players overall value and this is why I really dislike corsi analysis.  It completely ignores a significant, maybe the most significant, aspect of the game.  Furthermore, I believe that offensive ability and defensive ability should be evaluated separately, which many who do corsi analysis don’t do or only partially or subjectively do.

I really don’t know how many different ways I can show that shooting percentage matters a lot but there are still a lot of people who believe players can’t drive or suppress shooting percentage or believe that shooting percentage is a small part of the game that is dwarfed by the randomness/luck associated with it (which is only true if sample size is not sufficiently large).  The fact is corsi analysis alone will never give you a reliable (enough to make multi-million contract offers) evaluation of a players overall ability and effectiveness.  Shooting percentage matters, and matters a lot.  Ignore at your peril.

 

  14 Responses to “Showing Shooting Percentage Matters (Yet Again)”

  1.  

    David to the best of my knowledge nobody is denying that shooting percentage matters. You would look like less of an idiot if you tried to understand what others actually are saying than misrepresnting it.

    Shooting percentage on a team level in a differential measurement like Corsi is usually not very repeatable. It is best to neglect it on a low level analysis. It is included in a higher level analysis of course. Not in the way you want it included because you have made the mistake of neglecting Corsi which is far more important than shooting percentage and far more repeatable.

    In order to make your point, you often do dishonest things like mine quotes out of context (the Rob Vollman one) and ignore context which explains your examples (Drury vs. Gaborik). Worst of all you do not seem to understand the cocept of Corsi. You do not understand that its strength comes from the fact it is a differential measurement (shots for – shots against). This leads you to make nonsense statements about preventing shots in your own zone being an important part of scoring goals, which you did in the puck stops here thread you linked to.

  2.  

    “David to the best of my knowledge nobody is denying that shooting percentage matters.”

    Last week we spent 2 days exchanging comments where you downplayed the value of considering shooting percentage in player analysis. If it is important, we should consider it.

    “Shooting percentage on a team level…”

    I am not talking team level. I am talking player and line level. Big difference.

    “It is best to neglect it on a low level analysis. It is included in a higher level analysis of course.”

    Please explain the difference between low level and high level analysis.

    “Not in the way you want it included because you have made the mistake of neglecting Corsi which is far more important than shooting percentage and far more repeatable.”

    Clearly you haven’t read or understood anything I wrote above. Players can drive his on-ice shooting percentage. Driving on-ice shooting percentage is maybe the prime skill when it comes to being an elite offensive player. Gaborik isn’t good because he is good at helping his team generate shots, he is good because he helps his team capitalize on the shots they get.

    “…ignore context which explains your examples (Drury vs. Gaborik).”

    Here is your chance. Please provide some Drury vs Gaborik context. You spew a lot of wordy gibberish about who wrong I am but you never get around to actually providing me an example of how you would do things better. So please, provide me with your best Drury vs Gaborik analysis so I can get an idea of where you are coming from.

    “You do not understand that its strength comes from the fact it is a differential measurement (shots for – shots against).”

    How is that a strength? All a differential or ratio does is muddy the waters between offense and defense. With a differential, a good offense, bad defense player can have similar differentials as a bad offense, good defense player and yet the players have completely different skill sets and completely different value to their team. As I wrote above, I prefer to analyze offense and defense separately so I get a more complete picture of a players attributes and talents. Yes, there is value in knowing whether a player is net positive or net negative, but there is greater practical value in knowing whether the player is good offensively or good defensively.

    Finally, for someone who often likes to say “the numbers say this” or “the numbers say that” you seem to have an extreme adversity to actually using numbers to back up your claims. Maybe because the numbers don’t.

  3.  

    DJ: Last week we spent 2 days exchanging comments where you downplayed the value of considering shooting percentage in player analysis. If it is important, we should consider it.

    PSH: It is not nearly as important or as much of a skill as you think it is. It is not repeatable for the most part. Shooting percentage on a team level (by that I mean everyone on the ice on both teams when a given player is on the ice) is driven by the situation in the game. In a defensive situation it will be lower than in an offensive situation. The largest part of the repeatablity comes from players who are used in offensive situations one year being used in offensive situations the next year too.

    DJ: Please explain the difference between low level and high level analysis.

    PSH: A low level analysis is something quick and dirty where we are not interested in exact numbers. Merely we are interested in showing a general trend or that one player is better than another when there is a big difference between them. A high level analysis is when things have to be done with more accuracy.

    DJ: Please provide some Drury vs Gaborik context.

    PSH: Gaborik plays in offensive situations. Drury plays in defensive ones. Any team level shooting percentage difference between them is driven for the most part by the different situations in which they play.

    PSH: “You do not understand that its strength comes from the fact it is a differential measurement (shots for – shots against).”

    DJ: How is that a strength? All a differential or ratio does is muddy the waters between offense and defense

    PSH: Scorers bias is a big problem with hockey stats. By taking a differential measurement we do not run into problems that come from one scorer who credits missed shots far more often than another.

    Trying to answer offensive only or defensive only questions directly from Corsi analysis shows you don’t have a clue what the point to it is. The purpose is to show puck possession and puck position on the ice. You are asking an incoherent question and that is why you get answers that nobody looks at seriously.

    Your question is akin to asking What makes green such a violent rabbit. Its wrong because it makes no sense.

    You have incorrectly decided that Chris Drury can significantly change the shooting percentage of anyone who gets on the ice with him. The reality is Chris Drury played in defensive situations, so he and his linemates did not try to get inside (high percentage) shots. That one realization is the majority of your point. You have a stat that largely measures how offensive or defensive the role of a given player is and you wrongly think it is a skill.

    Have you ever played a team sport? If you had you would know that you cannot make any significant difference in the shooting percentage of a guy who happens to play alongside you.

  4.  

    Perhaps the best way to understand your error is by analogy. The situation a player plays in drives the shooting percentage of not on ly that player but also all the other players on the ice. You have mistaken cause and effect. The player does not drive things (at least not nearly on the level as the situation does). Certain players appear in different situations more than others. this will lead to differences between players that are largely driven by the situation in which the players play.

    It is like you arguing that the tree branches blowing cause wind. Proof is you did a lot of math to show that branches tend to blow on windy days. The problem is the windy days cause the branches to blow and not vice versa and your mah is thus wrong because you had incorrect assumptions that went into it. The situation drives the shooting percentage – you falsely attribute it to the player playing in the situation.

  5.  

    PSH: It is not nearly as important or as much of a skill as you think it is. It is not repeatable for the most part.

    DJ: But it is repeatable. I have shown that. You just need enough data to identify it.

    PSH: The largest part of the repeatablity comes from players who are used in offensive situations one year being used in offensive situations the next year too.

    DJ. You are making stuff up. I know for a fact that you haven’t done the analysis to show that.

    PSH: Gaborik plays in offensive situations. Drury plays in defensive ones. Any team level shooting percentage difference between them is driven for the most part by the different situations in which they play.

    DJ: So you are saying that the only difference between Gaborik’s 10.26% on-ice shooting percentage and Drury’s 6.18% on-ice shooting percentage is that Gaborik tries to score goals and Drury doesn’t?

    PSH: You have a stat that largely measures how offensive or defensive the role of a given player is and you wrongly think it is a skill.

    DJ: So I suppose that means that Gomez’s 6.34% on-ice shooting percentage means he primarily played a defensive role the past 4 seasons.

    PSH: If you had you would know that you cannot make any significant difference in the shooting percentage of a guy who happens to play alongside you.

    DJ: You could be right. Sidney Crosby has no ability to drive shooting percentage and no ability to make his line mates better. Furthermore, he ranks 52nd among forwards in 5v5 fenwick % over the past 4 seasons. What a bum. Can’t the Penguins do better for $8.7M? Maybe someone like Cheechoo (5th), or Moss (17th) or Fehr (20th), or even Zherdev (22nd) or Ponikarovsky (23rd). These guys are light years ahead of Crosby, we only need to ask them to play a more offensive style game they will be able to have >10% on-ice shooting percentages and they will be 80 point guys and best of all we can get 4 or 5 of them for what we are paying Crosby. Boy, this is going to be awesome.

    Then again, you could be wrong.

  6.  

    David

    Your responce is essentially this (but with a bit more math which is pointless because your underlying model is wrong).

    Wind is not driven by tree branches moving. It is the other way around. That said trees can affect the wind. They can be a wind block for a given area. They can form a wind tunnel if the wind gusts between the trees.

    If I truly want to know the exact details of the weather conditions I do need to know where trees are and wehat their effects are. That said I can do some important calculations without caring about the trees. Your weather forecast doesn’t include the trees. However if i am on the golf course trying to take a shot I do have to include the grove of trees beside the fairway to figure out wind conditions for my next shot. The problem is you insiste the trees are the most important thing to know wind conditions and that is outright stupid. You pull out quotes from others saying that the trees sheltered them from the winds as an argument. You carefully measure the wind beside your apple tree and show it is different from beside your pear tree. All of that doesn’t matter. Your whole premise is flawed because the winds cause the trees to move. The fact trees can also affect winds really is immaterial to your claim.

    Then you show that you don’t understand Corsi to try to knock it down. Your Sidney Crosby comment is so stupid and you know it is stupid. Crosby may not be the best player in the league terms of puck possession – that said he isn’t nearly as poor as you pretend by doing the stats wrong. Puck possession is not 100% of the game of hockey. You must know that. He ranks better by other measures.

    You think that you can take any player and play him in a different role and he will be equally good as he had been (that is your comment on Cheechoo, Moss etc.). That is stupid. It totally misses that their numbers (like all numbers) depend on context. You have generally picked as your examples players who could succeed in limited roles and were not asked to play a lot of minutes. Total numbers are better than per minute numbers for this reason. Do you want a 50 goal scorer on your team (in 80 games) or a 5 goal scorer in eight games? If it isn’t obvious the 50 goal scorer is the better choice you have no idea how to value players. That remains true even if the 50 goal scorer played 82 games and the 5 goal guy played 7 games to have a better goal per game.

  7.  

    PSH: Then you show that you don’t understand Corsi to try to knock it down. Your Sidney Crosby comment is so stupid and you know it is stupid. Crosby may not be the best player in the league terms of puck possession – that said he isn’t nearly as poor as you pretend by doing the stats wrong. Puck possession is not 100% of the game of hockey. You must know that. He ranks better by other measures.

    DJ: Ummm, yeah. I think that is the point I was trying to make. Puck possession doesn’t tell the whole story. If it did, Crosby wouldn’t be a superstar. He is a superstar because he converts puck possession into goals through a very good shooting percentage (one of the other measures).

    PSH: You think that you can take any player and play him in a different role and he will be equally good as he had been (that is your comment on Cheechoo, Moss etc.). That is stupid.

    DJ: No, that is not my argument, that was yours. Remember you are the one trying to suggest that puck possession is the best evaluation metric and shooting percentage is simply a result of playing style.

    PSH: Do you want a 50 goal scorer on your team (in 80 games) or a 5 goal scorer in eight games?

    DJ: That is an extreme example (not to mention 8 games is too small of a sample size to tell us anything), but I would take the guy who can score 40 goals in 15 minutes of ice time per game over the guy who can score 50 in 24 minutes of ice time per game because I can find someone who can get me the extra 10 goals, hopefully more, in the other 9 minutes. Or, I could find a way to get the 40 goal guy more minutes of ice time to improve his overall output.

  8.  

    David

    I am not arguing that Corsi is a single “be all and end all” statistic to evaluate players. It is one valuable number that goes with other valuable numbers. You are the one pushing that strawman with your Sidney Crosby example. That problem runs through your entire recent responce. It shows you have no idea what Corsi purports to measure and you think it fails because it doesn’t measure things it doesn’t claim to measure.

    •  

      I know what corsi measures. It measures puck control. What it doesn’t measure is the overall value of a player, or come anywhere close to it. My problem is not that you use corsi, it’s that you over emphasize its importance and under-emphasize the importance of other stats such as shooting percentage. And even more annoying is you do so without taking the time to really research the issue and instead are simply regurgitating the nonsensical drivel that guys like Gabe Desjardins spew.

      PSH: You are the one pushing that strawman with your Sidney Crosby example.

      DJ: Ahh yes, and this from the guy who initiated the conversation using an 8 game sample from an injured Brendan Morrison.

  9.  

    There is no stat that measures the overall value of a player. Complaining that a stat doesn’t do it is pretty pointless. All stats don’t. That makes it a stupid measuring stick.

    That said if I had two numbers to look at to evaluate a player – his Corsi in one season and his shooting percentage in one season and no other information – it would be intelligent to take the best Corsi player and neglect the shooting percentage numbers. That idea is obvious. You want the more reliable number that represents a more repeatable individualized skill.

    To carry on the wind/tree analogy, Corsi is equivlaent to the average climate in a region and shooting percentage the number of trees. Which do you think is more important if you wanted to predict which region might have the better weather next season?

    It is awfully stupid of you to think that my agreeing with Gabe Desjardins somehow is a negative. It seems everyone who looks at this problem winds up disagreeing with you and more or less agreeing with him. Why do you think that is?

    •  

      PSH: There is no stat that measures the overall value of a player. Complaining that a stat doesn’t do it is pretty pointless. All stats don’t. That makes it a stupid measuring stick.

      DJ: Ahhh, right. You are down to the “no stat is perfect, so don’t complain that this is not perfect” argument. The problem isn’t that is is not perfect, but that it gets misused and treated as being far more important than it is.

      PSH: That said if I had two numbers to look at to evaluate a player – his Corsi in one season and his shooting percentage in one season and no other information – it would be intelligent to take the best Corsi player and neglect the shooting percentage numbers. That idea is obvious. You want the more reliable number that represents a more repeatable individualized skill.

      DJ: Sure, in a world where I could only eat boiled brussels sprout and lima beans I might choose the lima beans. Thankfully we don’t live in that world, nor do we live in a world where we can only choose between corsi and shooting percentage. Your example is pointless. I’d choose goals.

      I came across this article yesterday: http://objectivenhl.blogspot.com/2011/03/loose-ends-part-i-predictive-validity.html

      “Corsi Tied is only marginally more predictive of future success than goal ratio or winning percentage when looking at samples of 60 games or more. In other words, as the sample size becomes increasingly large, there are diminishing returns with respect to the predictive advantage of Corsi. By the end of the season, all three variables seem to predict future success equally well”

      That is exactly the same as I came up with on the player “on-ice” level. At one year of data a goal analysis is about equal to corsi analysis as a predictor of future performance. JLikens didn’t take it any further, but beyond one year goal analysis is better.

      This is my #1 claim: If you want to evaluate a player, the best way to do so is a goal analysis on multiple seasons of data. If you have less than one year of data you can use corsi analysis but keep in mind that the results are still a somewhat poor predictor of future performance and thus a somewhat poor evaluation of player talent.

  10.  

    I don’t think its a very meaningful fact that goals become the best predictor of future success as the sample becomes large enough. The larger the sample the more luck will even out so if there is any talent at all, even the slightest bit, involved in scoring goals, that is separate from shot generation, that at some sample size goals will become the better predictor. This is not surprising and it doesn’t mean that shooting % is a more important facet of a player’s skill set than shot generation is. Goals is the combination of shot generation and shooting %. As the sample gets larger the volatility of shooting % becomes less of a problem and goals becomes the better predictor. So what? Shot generation is still the majority component of goal generation.

    •  

      You had me all the way until the last sentence. “Shot generation is still the majority component of goal generation.” The thing is, once you get a sample size large enough to minimize luck and stabalize shooting percentage, shooting percentage is more important in scoring goals than shot generation.

      •  

        I’d say this is what you need to do a better job of making a case for. But PSH’s contention that the situation is influencing the on ice shooting % more than the talent is something you’ll have to address. It will be difficult because you’ll need large samples and you’ll have to show that even when adjusted for situation, finishing talent has a greater influence on goal scoring than shot generation. I’m not sure you’ve proved this yet. Also, though, I’m not sure what exactly PSH means by situation. These #s are all 5v5, right? What other measurable factors can affect shooting %. The only thing I can think of is quality of opposition. I know the Sedins are well-known to play against weaker competition than the other lines (not sure how they manage this when they don’t have the last change). Also the game state (tied, w/in 1, down 2+, up 2+). It seems like a daunting task to create expected shooting %s based on situation and adjusted for competition and then compare them to what was observed.
        Then you’ll need to compare how much of a difference it makes in goal scoring totals compared to how much of a difference superior shot generation talent makes in goal scoring totals. Maybe there’s a shortcut. Good luck.

Sorry, the comment form is closed at this time.