[Simple Stats] Consistently Bad Defense in Losses

Discussion in 'Houston Rockets: Game Action & Roster Moves' started by yixiixiy, Feb 16, 2017.

  1. yixiixiy

    yixiixiy Member

    Joined:
    Feb 17, 2010
    Messages:
    854
    Likes Received:
    563
    I did some primitive data crunching. For all the games played this season so far, I measured the average and the standard deviation of opponent score in all 30 team's losses, and constructed the 95% confidence interval for each team's opponent score in a loss (meaning that, in the long run, there is 95% of the chance that opponent will score within this band in a team's loss).

    Here is the top (lowest score) and bottom (highest opponent score) 5.

    (Best defense in losses at the top):
    Rank Team Lower Upper
    1. Utah 98.3 105.8
    2. SA* 99.3 109.5
    3. Dallas 101.7 109.6
    4. Miami 102.1 109.4
    5. Detroit 103.7 110.4

    ...

    -5. Denver 112.1 118.6
    -4. Brooklyn 112.7 118.2
    -3. Houston 112.2 119.2
    -2. Phoenix 113.7 120.1
    -1 GS* 111.5 126.5

    [* GS and SA both lost too few games so far for the numbers to make statistical sense. Arguably Houston's 18 losses is also not sufficient to draw any meaningful inference. But one can perhaps see the trend and gain some insights here,]

    Also worth noting is that, for opponent scores in W's, Houston also ranked the 3rd worst, just below Brooklyn and Denver.

    This is somehow contradictory to the defensive rating on NBA.com and BB reference, which put us in the range of 10-15. I guess it's because those ratings are based on per 100 possessions, which could be somehow misleading as a means to standardize defensive performance. Houston plays a top 3 pace, but has mediocre rebounding, therefore giving the opponents more possessions. We let the other team score way more points than the defensive rating shows.

    We SUCK at defense.
     
    Houstunna, Ramo$e and heypartner like this.
  2. rocketsballin

    rocketsballin Member

    Joined:
    Nov 23, 2012
    Messages:
    7,449
    Likes Received:
    916
    thats weird, some pple here thing we're good enougn on defense to be contenders. you and i must be seeing a completely different rockets team then they are.
     
    fckbandwagons and yixiixiy like this.
  3. Lawlruschang

    Lawlruschang Member

    Joined:
    May 2, 2013
    Messages:
    157
    Likes Received:
    125
    I'd think missing capela/Beverley/nene for substantial time (strong rebounders at their positions, backups not so strong) have hurt our rebounding stats. At full strength I think we can even out the # of possessions issue. Good work though!
     
    Ramo$e likes this.
  4. heypartner

    heypartner Contributing Member

    Joined:
    Oct 27, 1999
    Messages:
    47,576
    Likes Received:
    14,233
    I don't understand this. Are you saying you prefer not doing per 100 possession. You need to standardize by Pace. If you didn't do that, then your numbers are flawed, imso. Sorry.

    You might be able to fix it real quick by grabbing the Pace number from either bball reference or NBA.com, and dividing final numbers by it. Accounting for Pace of the game is essential for these types of comparisons.

    thx for doing this. I like the idea of doing this band.
     
    Houstunna, yixiixiy and Ramo$e like this.
  5. FTW Rockets FTW

    FTW Rockets FTW Contributing Member

    Joined:
    Jun 23, 2011
    Messages:
    25,135
    Likes Received:
    15,240
    Not my fault. Go ask Jeff about the defense. I have nothing to do with it.

    - Mike AntoniO
     
  6. yixiixiy

    yixiixiy Member

    Joined:
    Feb 17, 2010
    Messages:
    854
    Likes Received:
    563
    Appreciate the comment. I guess I wasn't being clear.

    As a "per-unit" stat, defensive rating (per 100 possessions adjusted) is very transparent. And since it's standardized, it's valid comparing across teams using that. But eventually it's the # points lost per game that determines the defensive outcome of the game, and it's coming from multitude of things: per possession defensive efficiency, curbing opponent's pace, (to which rebounding, steal, and turnover all have contribution). Rockets' 10-15 ranked per possession defense is okay, but they need to do much more to allow fewer points.
     
  7. durvasa

    durvasa Contributing Member

    Joined:
    Feb 11, 2006
    Messages:
    29,073
    Likes Received:
    1,820
    OP, I don't understand your argument that using per possession stats is misleading for comparing across teams.
     
  8. yixiixiy

    yixiixiy Member

    Joined:
    Feb 17, 2010
    Messages:
    854
    Likes Received:
    563
    Please see my post right one floor above.
     
  9. heypartner

    heypartner Contributing Member

    Joined:
    Oct 27, 1999
    Messages:
    47,576
    Likes Received:
    14,233
    Whether true or not, you can't compare across teams without accounting for Pace....or at least using some sort of differential versus points giving up in their wins. I also think you might be introducing a notion of Volatility of teams between Wins and Losses, which is interesting. In other words, how much can a team control the tempo whether they win or not. But then we come back to efficiency is the key, and we lose that without Pace in the formula.

    wrt: why GSW is last. They are nowhere close to being worse defense. It's not just because of small sample size, it's because they will always show up *much* worse using Volume of Points vs using Pace to get you per possession stats.

    Can I ask a question here? I'm not sure we are talking about Possessions the same.

    Possessions alternate equally. Sounds like you are saying rebounds allow for extra possessions. No, they just reset the 24-second clock. The number of possession is roughly equal between opponents per game. Even if there are 5 offensive rebounds, that's still 1 possession according to definition of bball reference and NBA.com.... So each team will have the same Pace for each game....with slight variations depending upon outcome of tip and who has ball last per quarter.

    So, adjusting for Pace still preserves your concern about rebounds. If Rockets give up a lot of offensive rebounds, the opponents pts per possession will go up, because they get more chances per possession.
     
    yixiixiy and durvasa like this.
  10. durvasa

    durvasa Contributing Member

    Joined:
    Feb 11, 2006
    Messages:
    29,073
    Likes Received:
    1,820
    I agree with heypartner. The reason people use per-possession points allowed is that both teams will play almost exactly the same number of possessions. How good you are defensively is really just a matter of how well you limit the other team from scoring on each of their possessions.
     
    heypartner likes this.
  11. yixiixiy

    yixiixiy Member

    Joined:
    Feb 17, 2010
    Messages:
    854
    Likes Received:
    563

    I see. I think I didn't know exactly what a possession means.
     
  12. heypartner

    heypartner Contributing Member

    Joined:
    Oct 27, 1999
    Messages:
    47,576
    Likes Received:
    14,233
    Your definition is fine. TV analysts do talk in terms of offensive rebounds allowing more possessions. So, it has two definitions. The Statistics world defines them like football -- a possession always ends with a change of possession or end of quarter (end of half for football).

    Again, I like your idea of measuring a band/range of expected points. Comparing with the same band for wins, seems we might have a useful measure of Volatility. Rockets seem very volatile. If there is some measure of volatility between wins and losses, seems Vegas would be interested. I do know gamblers who talk in terms of volatility of scoring (mainly in NFL), such that they will shy away from betting on those games -- especially when choosing parlays and teasers.
     
    #12 heypartner, Feb 16, 2017
    Last edited: Feb 16, 2017
  13. PhiSlammaJamma

    Joined:
    Aug 29, 1999
    Messages:
    20,673
    Likes Received:
    1,206
    So you've verified their is some Defense being played. And Stuff. Better than I thought.
     
Loading...

Share This Page