Announcement

Collapse
No announcement yet.

What stats best capture players worth?

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

  • #16
    TeamEd wrote: View Post
    Well, for NetRtg you can discount Tyler and Chuck for small sample size. Next, you try to find an explanation for Kyle's lower than expected NetRtg. One that I think makes sense is that Kyle is logging a lot of minutes with TRoss and JV, two of the Raps' worst rotation players by on/off splits. Another is that Lou and JJ both play most of their minutes dominating opposing benches, so their NetRtg either slightly overstates their true value or correctly reflects that the bench is the Raps' greatest strength. Another might be that Kyle is excellent at collecting box score stats that inflate his advanced metrics like RPM and WAR beyond his true impact on the game. Finally, you account for the fact that Amir is a consistent on/off star.

    All of these explanations for Kyle's NetRtg (a still excellent +8) relative to the rest of the team are plausible, which is why you can't rely on one stat.
    Correct me if im wrong on this, but isn't Real Plus/Minus a top down statistic, meaning that it does not stem from the players box score? So, playing with others who have negative Net ratings or RPM should hurt Kyle Lowry's RPM? I may be wrong as I don't know the methodology for RPM and don't know how it is constructed.

    I see much better stats in the next few years with the addition of spatial tracking. I'm reading a book on it know which is very interesting.

    Comment


    • #17
      DogeLover1234 wrote: View Post
      Correct me if im wrong on this, but isn't Real Plus/Minus a top down statistic, meaning that it does not stem from the players box score? So, playing with others who have negative Net ratings or RPM should hurt Kyle Lowry's RPM? I may be wrong as I don't know the methodology for RPM and don't know how it is constructed.

      I see much better stats in the next few years with the addition of spatial tracking. I'm reading a book on it know which is very interesting.
      Yeah, ESPN's real plus minus is *not* a box score metric, it's a +/- stat that's adjusted for situational factors using formulas we don't actually know. I meant BBRef's statistical plus minus in that little note with on Kyle's NetRtg, that metric is based on the box score -- I think. It's all hard to follow.

      Since RPM adjusts for teammate strength, I think the best explanation for Kyle's near team average NetRtg is his minutes played with JV and TRoss against the opposition's top lineups.
      @EdTubb - edwardtubb at gmail

      Comment


      • #18
        stooley wrote: View Post
        Win shares and wins produced are fairly popular I believe. I'm on a phone now so can't elaborate, but you could probably google em
        Wins produced can be found on BoxScoreGeeks.com.

        They have recently been playing with some sort of adjustment to the standard Wins Produced (WP) to incorporate defense better. You can see it as ADJP48 on the 'Players' tab on their site; all other tabs show standard WP48 and WP.

        P.S. Here is the exact formula for Wins Produced, on the site of its inventor, Dave Berri. You can see these are a bunch of stats heads. For demonstration, here is their critique of RPM.
        Last edited by Kuh; Mon Dec 29, 2014, 06:24 PM.

        Comment


        • #19
          According to Wins Produced, Kyle Lowry is around the 10th best player in the NBA (as of time of writing), roughly tied with Damian Lillard.

          Comment


          • #20
            Wins produced is a very impressive stat, but in my mind at least it dramatically overvalues defensive rebounds - to the point of skewing the stat in a pretty extreme way. I find win shares fits more with typical valuations of players. RPM is very good - but it is important to remember it is a regressed stat over several years - it does not do well in capturing shorter term trends. It also has a big projection component - if you are looking to project a player's future performance it is pretty solidl but if you are looking to evaluate how they actually performed, the unregressed simple APM (as can be found on gotbuckets) is far better. There you will see more variance year to year, due to lack of regression, which is bad for projections but good because it does actually capture year to year changes better. It is still a rolling two-year average though, so keep that in mind.

            One thing I will say is to stay far away from basketball reference's box score plus minus. It is meant as a replication of APM, but it is built entirely off of box score stats (they looked to see which box score stats tend to correlate with better APM's and built relationships from that), so that they could be used to compare players historically. In other words, it is made intentionally less accurate than APM. Over small samples (before APM numbers are available) it can be useful, but it is in no way an equivalent or superior replacement for actual APM stats.
            twitter.com/dhackett1565

            Comment


            • #21
              DanH wrote: View Post
              Wins produced is a very impressive stat, but in my mind at least it dramatically overvalues defensive rebounds - to the point of skewing the stat in a pretty extreme way.
              This is the main reason it doesn't pass the sniff test for me. Tyson Chandler, DeAndre Jordan, Brandan Wright, and Rudy Gobert are all in the top 10 of WP/48 this year. By way of comparison, LeBron is 37th.

              As others have said, every stat requires context, but I like using a mix of TS%, win shares, and PER to get a rough idea of how a player stacks up, since no one stat tells you the whole story. They each have their own flaws.

              Comment


              • #22
                DogeLover1234 wrote: View Post
                Correct me if im wrong on this, but isn't Real Plus/Minus a top down statistic, meaning that it does not stem from the players box score? So, playing with others who have negative Net ratings or RPM should hurt Kyle Lowry's RPM? I may be wrong as I don't know the methodology for RPM and don't know how it is constructed.

                I see much better stats in the next few years with the addition of spatial tracking. I'm reading a book on it know which is very interesting.
                Which book is it? I'd love to read it


                Sent from my iPhone using Tapatalk
                A key that opens many locks is a master key, but a lock that gets open by many keys is just a shitty lock

                Comment


                • #23
                  3inthekeon wrote: View Post

                  BRef's Win Shares seems more a group than individual stat. A poor defender playing with good defenders will get positive Defensive Win Shares for instance.
                  I agree, but basketball is a team game. The way you work in the system could make the sum greater than the parts. Amir Johnson for instance. I haven't found a stat for screen setting(apart from one vantage sports article), but I'd bet he is one of the better ones. That, and all the other dirty work, shows up in the on/off tab on BRef. I especially like the overall eFG% stat. shows how a player helps the offence and defence.

                  Comment


                  • #24
                    rus wrote: View Post
                    I agree, but basketball is a team game. The way you work in the system could make the sum greater than the parts. Amir Johnson for instance. I haven't found a stat for screen setting(apart from one vantage sports article), but I'd bet he is one of the better ones. That, and all the other dirty work, shows up in the on/off tab on BRef. I especially like the overall eFG% stat. shows how a player helps the offence and defence.
                    Well RPM was designed to pick up players overall worth, which is why Amir has been a Top 40 player in the NBA this year and last by that metric. Win Shares takes "team game" to an extreme. Carlos Boozer in no way helps a teams defence, but got 4.3 Defensive Win Shares last year by playing a lot with Noah, Butler, and the other Bulls who are plus defenders.

                    Also could you explain how eFG% helps defence?
                    If we knew half as much about coaching an NBA team as we think, we"d know twice as much as we do.

                    Comment


                    • #25
                      Scraptor wrote: View Post
                      This is the main reason it doesn't pass the sniff test for me. Tyson Chandler, DeAndre Jordan, Brandan Wright, and Rudy Gobert are all in the top 10 of WP/48 this year. By way of comparison, LeBron is 37th.
                      The big thing about WP is that it penalises players for missing shots (they recently added a correction for defensive rebounds).

                      Tyson, DeAndre, Brandan and Rudy are shooting 68%, 70%, 73% and 63% respectively. That's why those four are so highly ranked. Jonas Valanciunas, who rebounds at a comparable rate to the last two, but is shooting only 54%, ranks much further down.

                      Re Lebron: it's worth pointing out that last year LeBron ranked 3rd in WP/48, just marginally behind Chris Paul and Andre Drummond.

                      Maybe the fact that LeBron is 37th (and Drummond is 70th or so) means he's not playing as well as last year?

                      In fact if you look at LeBron's stats over the years, you will see that he has been consistently a great performer according to WP, but he has only been transcendent superstar great in 5 seasons: his last two in Cleveland, and his last three in Miami. Before/between those years his performance dropped significantly: i.e., to Tim Duncan levels rather than LeBron levels.

                      Comment


                      • #26
                        e_wheazhy_ wrote: View Post
                        Which book is it? I'd love to read it


                        Sent from my iPhone using Tapatalk
                        Agreed lol. Can we get a title?
                        "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


                        • #27
                          What stats best capture players worth?

                          stooley wrote: View Post
                          Agreed lol. Can we get a title?
                          Patience my friend, our title will come. For now I just want to know the name of the book
                          A key that opens many locks is a master key, but a lock that gets open by many keys is just a shitty lock

                          Comment


                          • #28
                            Kuh wrote: View Post
                            The big thing about WP is that it penalises players for missing shots (they recently added a correction for defensive rebounds).

                            Tyson, DeAndre, Brandan and Rudy are shooting 68%, 70%, 73% and 63% respectively. That's why those four are so highly ranked. Jonas Valanciunas, who rebounds at a comparable rate to the last two, but is shooting only 54%, ranks much further down.

                            Re Lebron: it's worth pointing out that last year LeBron ranked 3rd in WP/48, just marginally behind Chris Paul and Andre Drummond.

                            Maybe the fact that LeBron is 37th (and Drummond is 70th or so) means he's not playing as well as last year?

                            In fact if you look at LeBron's stats over the years, you will see that he has been consistently a great performer according to WP, but he has only been transcendent superstar great in 5 seasons: his last two in Cleveland, and his last three in Miami. Before/between those years his performance dropped significantly: i.e., to Tim Duncan levels rather than LeBron levels.
                            WS has an almost identical relationship with missed shots as WP does - they differ in rebounding mostly.

                            It is PER that has a backwards relationship with missed shots - high usage, low efficiency guys look good in PER and bad in the wins stats.
                            twitter.com/dhackett1565

                            Comment


                            • #29
                              Dan, earlier you mentioned RPM is a regressed stat over several year, yet that's not how I read it, and the BR story seems to confirm each seasons stats are unique to that season.

                              Here is the BR link.
                              http://bleacherreport.com/articles/2...how-they-do-it
                              If we knew half as much about coaching an NBA team as we think, we"d know twice as much as we do.

                              Comment


                              • #30
                                3inthekeon wrote: View Post
                                Dan, earlier you mentioned RPM is a regressed stat over several year, yet that's not how I read it, and the BR story seems to confirm each seasons stats are unique to that season.

                                Here is the BR link.
                                http://bleacherreport.com/articles/2...how-they-do-it
                                Here's your mistake - you are reading bleacher report.

                                RPM is an updated version of xRAPM - which means regularized adjusted plus minus. If you look into the history of RAPM, you'll see it is very much a multi-year stat.
                                twitter.com/dhackett1565

                                Comment

                                Working...
                                X