Announcement

Collapse
No announcement yet.

Quantifying Shot Quality in the NBA - Sloan Conference paper

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Quantifying Shot Quality in the NBA - Sloan Conference paper

    Interesting read from the Sloan Conference on trying to quantify shot quality. Very technical stuff and I'm still trying to wrap my head around it.

    According to the data, Gerald Green is the best mid-range shooter in the game (Demar isn't mentioned in their top 5). Jose Calderon is the 2nd best overall, behind Kyle Korver. This isn't to say they are the best shooters, but rather they take the best shots and make them at a high rate.

    http://www.sloansportsconference.com...in-the-NBA.pdf
    Heir, Prince of Cambridge

    If you see KeonClark in the wasteland, please share your food and water with him.

  • #2
    ''Very technical stuff''
    loool you don't say :P

    No DeMar??? kinda hard to believe, I'll take your word for it tho

    Jose?? who is this Jose whom you speak of?
    What they got to say now? Nothing they can say now. Mobbin' on the low. Winnin' on the low
    The city embraced me, made me feel at home. The only difference [between Compton and Toronto] for me is the cold. -DeMar
    No Where Near the South Side #WeTheNorth

    Comment


    • #3
      Interesting stuff. I love to see more context being built into these metrics.
      "Bruno?
      Heh, if he is in the D-league still in a few years I will be surprised.
      He's terrible."

      -Superjudge, 7/23

      Hope you're wrong.

      Comment


      • #4
        Here's the list of all the papers from the Conference.

        http://www.sloansportsconference.com/?page_id=462
        Heir, Prince of Cambridge

        If you see KeonClark in the wasteland, please share your food and water with him.

        Comment


        • #5
          stooley wrote: View Post
          Interesting stuff. I love to see more context being built into these metrics.
          In time, I believe we will.

          One of the better (and easier to comprehend) things from the paper:
          "This type of approach where we separate shooting into shot quality and shooting ability also allows us to more properly value players whose EFGs are undercut by shot quality. A great example of this is Dirk Nowitzki, who ranked 64th in our list with an EFG of 53.9. However, Dirk had the 6th lowest ESQ (44.0) meaning he has been taking very difficult shots. Adjusting for this, Dirk has an EFG+ of 10.0, which is 6th in the league and more reflective of his elite shooting ability."

          Some players are elite shooters, others are just getting easier shots, whether by their teammates or their own decisions is to be left to the eye test or future data. Would love to see the full list of players and see how the Raps ranked.
          Last edited by Axel; Mon Mar 3, 2014, 03:07 PM.
          Heir, Prince of Cambridge

          If you see KeonClark in the wasteland, please share your food and water with him.

          Comment


          • #6
            Axel wrote: View Post
            In time, I believe we will.

            One of the better (and easier to comprehend) things from the paper:
            "This type of approach where we separate shooting into shot quality and shooting ability also allows us to more properly value players whose EFGs are undercut by shot quality. A great example of this is Dirk Nowitzki, who ranked 64th in our list with an EFG of 53.9. However, Dirk had the 6th lowest ESQ (44.0) meaning he has been taking very difficult shots. Adjusting for this, Dirk has an EFG+ of 10.0, which is 6th in the league and more reflective of his elite shooting ability."

            Some players are elite shooters, others are just getting easier shots, whether by their teammates or their own decisions is to be left to the eye test or future data. Would love to see the full list of players and see how the Raps ranked.
            I can understand the paper pretty well - I'm used to economics papers (which may actually be less of a hard science than basketball).

            I love the separation into ESQ and EFG+ - quite interesting. I'd imagine there's slightly more to shot difficulty than just location, the closest defender and whether or not it was off the dribble, but I think that paints a pretty darn good picture.
            "Bruno?
            Heh, if he is in the D-league still in a few years I will be surprised.
            He's terrible."

            -Superjudge, 7/23

            Hope you're wrong.

            Comment


            • #7
              One thing that would be interesting to carry this into would be the effect of taking higher or lower ESQ on the team as a whole.

              How does Lamarcus Aldridge's shot selection affect future shots by his teammates.

              Also, it would be nice to see ESQ for various types of shots - post ups, floaters and hard drives for example - or for each type of player - at each position or something like that.
              Last edited by stooley; Mon Mar 3, 2014, 03:56 PM.
              "Bruno?
              Heh, if he is in the D-league still in a few years I will be surprised.
              He's terrible."

              -Superjudge, 7/23

              Hope you're wrong.

              Comment


              • #8
                Just thought of this issue -

                The paper is comparing each player's shots against other identical shots - but not every player takes the same type of shots.
                This means the ESQ base will be skewed by whatever sample of players that have taken that shot, and at what rate they're each taking it - it is not necessarily the average player's ability to make that shot.

                For example, if a bad shooter, say josh smith, is taking a very large number of off-the-dribble long twos from a specific point on the baseline, where not THAT many players shoot from, then that section's ESQ base will be skewed low.

                In section 2 of the paper, where they show a chart of the half-court with the small areas - the percentages look a little wonky as you move from the left side of the baseline into the basket, for example.
                Last edited by stooley; Mon Mar 3, 2014, 03:54 PM.
                "Bruno?
                Heh, if he is in the D-league still in a few years I will be surprised.
                He's terrible."

                -Superjudge, 7/23

                Hope you're wrong.

                Comment


                • #9
                  stooley wrote: View Post
                  Just thought of this issue -

                  The paper is comparing each player's shots against other identical shots - but not every player takes the same type of shots.
                  This means the ESQ base will be skewed by whatever sample of players that have taken that shot, and at what rate they're each taking it - it is not necessarily the average player's ability to make that shot.

                  For example, if a bad shooter, say josh smith, is taking a very large number of off-the-dribble long twos from a specific point on the baseline, where not THAT many players shoot from, then that section's ESQ base will be skewed low.

                  In section 2 of the paper, where they show a chart of the half-court with the small areas - the percentages look a little wonky as you move from the left side of the baseline into the basket, for example.
                  True, but the data can only capture reality. It's not like we can line up all the various players and force them to take specific shots for the data feed. While it's possible that certain players would skew certain shot types, I don't think that is any difference than any league wide stat (league average FG% for a position for example would include chuckers like JR Smith); and any outlier can be easily identified.

                  If Josh Smith is taking a disproportionate amount of off the dribble long twos, then that volume is noted, his Shot Quality score plummets. Once it is identified, removing a single player's stats to compare the rest of the league is a fairly simple calculation; that would give you both the true score (league wide with all players), the Josh Smith score (his own individual stat) and the adjusted league score (league wide minus Josh Smith). That would give you 3 data points that can be compared fairly.

                  To the bold, I think the overall point is, in part, to demonstrate which players take the best shots. Durant takes a lot of harder shots, so his ESQ declines, but his EFG+ remains high because of his elite talent. Similar to how Dirk scored low due to the difficulty of his shots, but once the margin of differential was calculated, he ranked 6th best because of his elite skill. The next step for these players is to try and take "easier" shots, which, combined with their elite skill, would propel their efficiency. Dirk might have a harder time with that due to the skill level of his teammates, but Durant has Westbrook and Ibaka; perhaps Scott Brooks needs to re-tool the Thunder offence to get his elite players in position for better shots. The fact that the Thunder could be better by schematic design is a scary thought.

                  Based on the Raps eye test (since we don't have their data), I'd have to guess that Ross has a fairly high ESQ since the majority of his shots are the catch and shoot variety. If he can retain that level of shot quality, and increase his skill (practice, practice, practice) then he should be a very efficient scorer. This then gets into the paradox of increased scoring means increased defensive attention and therefore more difficult shots (can we just call this The Demar Factor?). I think that is one of the reasons why I've personally been so critical of Demar, his shot quality has never seemed particularly great. But a large part of that could be attributed to the quality of his teammates. I wonder how Demar would have been viewed if him and Ross had been reversed, with Ross being the more experienced floor stretcher and a rookie Demar working in the spaces provided. Instead, Demar didn't have the space to operate in his "preferred" area causing his shot quality to decrease. If that had been switched, would we have viewed Demar as an efficient player who was simply "working off his teammate" as opposed a player who was forcing the play into the inefficient mid-range game? Perception often becomes reality, and it seems plausible that there would have been significant difference in how he'd be viewed if that was the case (perhaps Matt52 and I would never have been called 'haters' in this alternate reality either ).

                  I think the other area that this data can help identify is in regards to team ball movement and big man FG%. Some bigs have sky high FG% because of the quality of their looks (think DeAndre Jordan with the Clippers). When comparing two bigs as "who's the better scorer down low", consideration has to be given to the fact that some players get wide open looks largely due to their team's better ball movement, which others are given the ball on the block and asked to score. Perfect comparison for example would be Jordan vs Al Jefferson. Jefferson would not receive the same benefits from his Charlotte teammates as Jordan would from Chris Paul and Blake Griffin (and JJ Reddick and etc). So while Jordan would presumably have a very high FG% (it's 66.5%), his shots should be significantly easier than that of Al Jeff (49.7%). Al Jeff would have a lower ESQ but is clearly the better scorer. When comparing two lesser know entities, this could be a valuable tool to find "diamonds in the rough". I'd be very curious to see the ESQ of JV, Amir and PP since they'd all have the same teammate benefit factor.
                  Heir, Prince of Cambridge

                  If you see KeonClark in the wasteland, please share your food and water with him.

                  Comment


                  • #10
                    Axel wrote: View Post
                    True, but the data can only capture reality. It's not like we can line up all the various players and force them to take specific shots for the data feed. While it's possible that certain players would skew certain shot types, I don't think that is any difference than any league wide stat (league average FG% for a position for example would include chuckers like JR Smith); and any outlier can be easily identified.

                    If Josh Smith is taking a disproportionate amount of off the dribble long twos, then that volume is noted, his Shot Quality score plummets. Once it is identified, removing a single player's stats to compare the rest of the league is a fairly simple calculation; that would give you both the true score (league wide with all players), the Josh Smith score (his own individual stat) and the adjusted league score (league wide minus Josh Smith). That would give you 3 data points that can be compared fairly.
                    I was just nitpicking at how this stat isn't quite perfect. On the surface, it seems to paint a very accurate picture. I was just looking for small assumptions or errors in the manipulation of the data to see where perhaps this stat is flawed. We all know WS overvalues 3rd options, for example.

                    For example, let's imagine that there's a point on the baseline, about 6 or 7 feet out, where the vast majority of shots are turn around left handed baby-hooks taken by centers. The player then, who drives, facing the basket, and pulls up for a bank shot, is perhaps taking a shot with a higher ESQ than what this system attributes to it.

                    So, to comment on your Deandre Jordan/Al Jefferson analysis - this stat would likely not take into account how each player receives the ball. ESQ looks at the nearest player, the distance from the basket, and the presence of dribbles.

                    When Deandre Jordan's man moves in front of him to hedge on a Blake Griffin face up, and Deandre receives a pass over his man, puts it on the ground for a dribble and dunks the ball with no one between him and the basket, but with his man trailing him by 2 feet, he receives a certain ESQ.

                    That ESQ would be identical to Al Jefferson receiving a pass in the post, backing his man down and making a much more difficult post-move, so long as the nearest defender is equal distance as to Blake Griffin.

                    But Deandre Jordan likely makes his shot far more often, so this skews his EFG+ up.

                    But I do like this stat a lot. I wish we could somehow access information like how far the closest defender was.
                    "Bruno?
                    Heh, if he is in the D-league still in a few years I will be surprised.
                    He's terrible."

                    -Superjudge, 7/23

                    Hope you're wrong.

                    Comment


                    • #11
                      stooley wrote: View Post
                      I was just nitpicking at how this stat isn't quite perfect. On the surface, it seems to paint a very accurate picture. I was just looking for small assumptions or errors in the manipulation of the data to see where perhaps this stat is flawed. We all know WS overvalues 3rd options, for example.

                      For example, let's imagine that there's a point on the baseline, about 6 or 7 feet out, where the vast majority of shots are turn around left handed baby-hooks taken by centers. The player then, who drives, facing the basket, and pulls up for a bank shot, is perhaps taking a shot with a higher ESQ than what this system attributes to it.

                      So, to comment on your Deandre Jordan/Al Jefferson analysis - this stat would likely not take into account how each player receives the ball. ESQ looks at the nearest player, the distance from the basket, and the presence of dribbles.

                      When Deandre Jordan's man moves in front of him to hedge on a Blake Griffin face up, and Deandre receives a pass over his man, puts it on the ground for a dribble and dunks the ball with no one between him and the basket, but with his man trailing him by 2 feet, he receives a certain ESQ.

                      That ESQ would be identical to Al Jefferson receiving a pass in the post, backing his man down and making a much more difficult post-move, so long as the nearest defender is equal distance as to Blake Griffin.

                      But Deandre Jordan likely makes his shot far more often, so this skews his EFG+ up.

                      But I do like this stat a lot. I wish we could somehow access information like how far the closest defender was.
                      I assumed that the distance of the defender would have only captured defenders in front of (or in the periphery of) the offensive player; since a defender trailing would have negligible impact. But now that you point that out, I'm not sure.

                      For your example, the data exists that they could differentiate between the pull-up jumper and the bank shot, as the player tracker data on NBA.com already distinguishes between drives (starting 20 feet out or something) and catch and shoot variables. Since the data exists, the ESQ could take that into account and may already. Technically speaking, a drop step baby hook would likely be considered more catch and shoot then off the dribble, so there should already be different coding for it. Again, all hypothetical since we don't have full access.

                      Ever notice that we seem to be just talking to each other on this stuff? lol
                      Heir, Prince of Cambridge

                      If you see KeonClark in the wasteland, please share your food and water with him.

                      Comment


                      • #12
                        Axel wrote: View Post
                        Ever notice that we seem to be just talking to each other on this stuff? lol
                        I have noticed that actually. Hoping that we don't buzzkill the whole forum.
                        "Bruno?
                        Heh, if he is in the D-league still in a few years I will be surprised.
                        He's terrible."

                        -Superjudge, 7/23

                        Hope you're wrong.

                        Comment


                        • #13
                          stooley wrote: View Post
                          I have noticed that actually. Hoping that we don't buzzkill the whole forum.
                          Some of us just read

                          Comment


                          • #14
                            stooley wrote: View Post
                            I have noticed that actually. Hoping that we don't buzzkill the whole forum.
                            Can't see it being a buzzkill since people choose to read a thread or not. Personally, I like this type of discussion as it can help us understand the future conversations relating to scouting, player development and talent assessment. Also could help devise an ideal coaching schematic.

                            At the end of the day, there are only so many conversations that can be had about how the last game went, or whether Lowry will be back next year. Bigger ideas are what gets you through the season sometimes.

                            Glad that my 2,000th post went to a worthwhile topic.
                            Heir, Prince of Cambridge

                            If you see KeonClark in the wasteland, please share your food and water with him.

                            Comment


                            • #15
                              Axel wrote: View Post
                              Can't see it being a buzzkill since people choose to read a thread or not. Personally, I like this type of discussion as it can help us understand the future conversations relating to scouting, player development and talent assessment. Also could help devise an ideal coaching schematic.

                              At the end of the day, there are only so many conversations that can be had about how the last game went, or whether Lowry will be back next year. Bigger ideas are what gets you through the season sometimes.

                              Glad that my 2,000th post went to a worthwhile topic.
                              Completely agree. Congrats on the 2000th post.
                              "Bruno?
                              Heh, if he is in the D-league still in a few years I will be surprised.
                              He's terrible."

                              -Superjudge, 7/23

                              Hope you're wrong.

                              Comment

                              Working...
                              X