1. Welcome! Please take a few seconds to create your free account to post threads, make some friends, remove a few ads while surfing and much more. ClutchFans has been bringing fans together to talk Houston Sports since 1996. Join us!

An Analytical Approach to the Rockets 2013-2014 season

Discussion in 'Houston Rockets: Game Action & Roster Moves' started by jtr, Sep 18, 2013.

Tags:
Page 6 of 6
  1. JustAGuy

    JustAGuy Member

    Joined:
    Dec 17, 2012
    Messages:
    1,464
    Likes Received:
    70
    That looks pretty complete, though I'm not sure which fields can be manipulated. Also, which spreadsheet fields are you using for the wins calculation (e.g. using the age field, or just displaying it?).
     
    #101 JustAGuy, Sep 24, 2013
  2. jtr

    jtr Member

    Joined:
    Dec 4, 2011
    Messages:
    7,470
    Likes Received:
    275
    I am just displaying all of the fields in the top list box.

    The fields used for team wins calculated are Win Share, Minutes Played, a constant of total minutes per position for the season (an approximation because I am not taking into account the overtime minutes played last season, but it is close), and a normalized number of the entire Minutes Played for the roster next season. I am not looking at the calculation so it may include more fields.

    For individual players it is just their gross Win Share not modified by minutes played.
     
    #102 jtr, Sep 24, 2013
    Last edited: Sep 24, 2013
  3. durvasa

    durvasa Member

    Joined:
    Feb 11, 2006
    Messages:
    38,893
    Likes Received:
    16,449
    This is going to be based on 2012/2013 stats, correct?

    Any plans to update it with upcoming season stats?
     
    #103 durvasa, Sep 24, 2013
  4. jtr

    jtr Member

    Joined:
    Dec 4, 2011
    Messages:
    7,470
    Likes Received:
    275
    Yes. But the tricky question is when to update. How long into the 2013-2014 season until the sample size becomes significant?
     
    #104 jtr, Sep 24, 2013
  5. Convictedstupid

    Convictedstupid Member

    Joined:
    Jul 5, 2008
    Messages:
    4,963
    Likes Received:
    512
    Everyone should read your post JCDenton. Brilliant. I laughed.
     
    #105 Convictedstupid, Sep 24, 2013
  6. Convictedstupid

    Convictedstupid Member

    Joined:
    Jul 5, 2008
    Messages:
    4,963
    Likes Received:
    512
    29ish games, right?
     
    #106 Convictedstupid, Sep 24, 2013
  7. jtr

    jtr Member

    Joined:
    Dec 4, 2011
    Messages:
    7,470
    Likes Received:
    275
    I don't know. I am going to have to look at sigma to determine how far is is from last seasons sigma before I make a decision.

    But 29 games sounds about right.
     
    #107 jtr, Sep 24, 2013
  8. Convictedstupid

    Convictedstupid Member

    Joined:
    Jul 5, 2008
    Messages:
    4,963
    Likes Received:
    512

    Obviously you want 1000s of samples but I think at around 29 is when samples become statistically significant.

    But I'm basing this off whether I remember whatever my prof said in stats 2 years ago so idk.
     
    #108 Convictedstupid, Sep 24, 2013
  9. JustAGuy

    JustAGuy Member

    Joined:
    Dec 17, 2012
    Messages:
    1,464
    Likes Received:
    70
    OK. Here's my quicky UI approach, keeping in mind that I am not familiar with what is possible/easy with tableau and I haven't thought about it much...

    but if I were writing a per team UI I'd have 3 sliders per player (xrapm & win share & minutes).. win share would only slide in a reasonable range for the player (+/- 2 ?). Ditto xrapm. Sliding minutes is the hard part, obviously, as total minutes must add up. I'd let the user handle that conundrum by not allowing sliders to increase unless team minutes are availabe, and not displaying results until 0 minutes available (with suitable message like "minutes must add up to 19680 or something like that). None of that crazy hard automagically decreasing someone's minutes when someone else is increased....

    There would be "calculators" available above that to help users-
    1) minutes calculator with 2 sliders, 1 minutes per game, 1 injured days. Tells you what the output minutes are given your assumptions on those two items. Selecting a player automatically populates it with that player's totals for prev. year
    2) inverse of above for roll players forced to fill in when starter is injured (3 sliders- normal minutes, replacement days, extra minutes per replacement day)
    3) W/S calculator which allows you to manipulate a couple/few player fields and spits out a new w/s for that player... selecting a player automatically populates it with that players totals from previous year.
    4) can't really make an xrapm calculator, though if you are breaking things out by offensive and defensive for win calculation i suppose that could matter...

    I'd chop everything else out (visually that is...)

    I can't tell to what extent you already have the sliders for w/s & xrapm (are those dots movable?).

    Any way, one man's opinion. Would be fun to play with, but I'm not gonna build it.
     
    #109 JustAGuy, Sep 24, 2013
  10. jtr

    jtr Member

    Joined:
    Dec 4, 2011
    Messages:
    7,470
    Likes Received:
    275
    I would not want to build what I am building, but I am doing it anyway. I will sit down in the next day or so with what you have suggested and figure out what what would be involved.

    Thanks for your input!
     
    #110 jtr, Sep 24, 2013
  11. jtr

    jtr Member

    Joined:
    Dec 4, 2011
    Messages:
    7,470
    Likes Received:
    275
    Having spent a little bit of time looking at it the problem is xRAPM. xRAPM has an obvious flaw in that even at seasons end some of the lineups analyzed are not statistically significant. I will have to look into it in greater depth.
     
    #111 jtr, Sep 24, 2013
  12. Convictedstupid

    Convictedstupid Member

    Joined:
    Jul 5, 2008
    Messages:
    4,963
    Likes Received:
    512
    Trades at the deadline and rookies finally breaking into the lineup during the 2nd half of the season would muddle the data a bit.
     
    #112 Convictedstupid, Sep 24, 2013
  13. jtr

    jtr Member

    Joined:
    Dec 4, 2011
    Messages:
    7,470
    Likes Received:
    275
    At this time I am really struggling to get good data for next season. It is all about quality data. I still do not have valid roster data for CLE next season. Cr**. I have spent three times as much time on data than I have spent on programming.
     
    #113 jtr, Sep 24, 2013
  14. JustAGuy

    JustAGuy Member

    Joined:
    Dec 17, 2012
    Messages:
    1,464
    Likes Received:
    70
    Just assume Lebron is on CLE. Problem solved.
     
    #114 JustAGuy, Sep 25, 2013
  15. JustAGuy

    JustAGuy Member

    Joined:
    Dec 17, 2012
    Messages:
    1,464
    Likes Received:
    70
    And yes, getting data is always a bigger pain than anything else. And then you have to skew your code based on the available data.
     
    #115 JustAGuy, Sep 25, 2013
  16. jtr

    jtr Member

    Joined:
    Dec 4, 2011
    Messages:
    7,470
    Likes Received:
    275
    I do not skew the code. If what I program does not make sense then it is either the code or the data.

    Take this view. The xRAPM scores on the right add up to something less than 42 wins per team. The WS adds up to 44 wins per team. I will accept the xRAPM total as within the margin of error. The WS is off by 8%. Not acceptable. Math or data.

    [​IMG]
     
    #116 jtr, Sep 25, 2013
  17. JustAGuy

    JustAGuy Member

    Joined:
    Dec 17, 2012
    Messages:
    1,464
    Likes Received:
    70
    By skew the code I'm referring to my own current process... it's different as I'm looking less to solve a problem and more for an interesting view of the data. When the data sucks, I change what I'm looking for/at. I'm still in the exploratory phase. Not a technique you can follow.
     
    #117 JustAGuy, Sep 25, 2013
  18. jtr

    jtr Member

    Joined:
    Dec 4, 2011
    Messages:
    7,470
    Likes Received:
    275
    Well I am working on various views.

    Salary for all NBA centers above ~7.5M
    [​IMG]
    Sallary for every Rocket player who has played a season in the NBA
    [​IMG]

    And here is one that shows a bubble view of all players with a positive RAPM and WS, but means nothing as far as I can tell. But it sure is pretty:
    [​IMG]
     
    #118 jtr, Sep 26, 2013
  19. JustAGuy

    JustAGuy Member

    Joined:
    Dec 17, 2012
    Messages:
    1,464
    Likes Received:
    70
    So is that last one WS/$ for radius (or something like that)?
     
    #119 JustAGuy, Sep 26, 2013
  20. jtr

    jtr Member

    Joined:
    Dec 4, 2011
    Messages:
    7,470
    Likes Received:
    275
    It looks to me like the radius is the sum of WS and RAPM.
     
    #120 jtr, Sep 26, 2013
(You must log in or sign up to post here.)
Page 6 of 6

Share This Page