OT: "High Tech" Hoops stats

Discussion in 'Chicago Bulls' started by transplant, Jan 28, 2011.

  1. transplant

    transplant Global Moderator Staff Member Global Moderator

    Joined:
    Jan 6, 2011
    Messages:
    4,111
    Likes Received:
    210
    Trophy Points:
    63
    I'd like your thoughts on some stats-related stuff. A couple reasons:

    - I recently got into a bit of a debate on another board regarding the DRTG stat shown on basketball-reference.com as it is applied to individuals. Apparently, the formula is so complex that you have to buy a book to review it. My POV is that, if I don't know what goes into a stat, I can't put any credence in it. Also, the short-form description of individual DRTG I've most often heard is that it essentially takes the team's DRTG and then assigns individual DRTGs on the basis of box score info (defensive rebounds, steals and blocks). That just strikes me as worthless. Agree or disagree?

    - As you may know, I've begun writing articles for sportstwo.com. I'm thinking about taking a shot at writing a series on some of the more advanced stats (TS%, PER, etc) targeted to the casual fan...an attempt to both introduce and demystify these numbers.

    I realize that almost all stats are good on their own...it's how they're used by folks that most often causes problems.

    In any case, what are your general feelings about the "new generation" of basketball statistics? Which do you think are the most interesting/useful? The most misunderstood and/or misused? Also, do you have any good stats sources other than basketball-reference and 82games?

    Thanks in advance for your help.
     
  2. MikeDC

    MikeDC Member

    Joined:
    May 24, 2007
    Messages:
    5,643
    Likes Received:
    16
    Trophy Points:
    38
    Occupation:
    Professor
    Location:
    Indianapolis, IN
    I'll write you a better reply tonight when I have the chance, but one resource you should check out is basketballvalue.com. They calc out adjusted plus minus, which is useful but pretty noisy (it varies a lot based on time and situation).

    I think your intuition about DRtg is exactly right. For individual players, especially non big men, it's largely useless.

    I'd warn against saying that you need to fully understand everything to give it much credence. I'd rely more on results. We may not understand how our cars, or turbotax works, or how our credit scores get calculated, but that doesn't mean we should ignore the info and benefits they provide.
     
  3. transplant

    transplant Global Moderator Staff Member Global Moderator

    Joined:
    Jan 6, 2011
    Messages:
    4,111
    Likes Received:
    210
    Trophy Points:
    63
    Thanks, Mike.

    Just to be clear, I'm not saying I need to be able to calculate the stat to give it credence, but I do need to understand the dynamics of it...what its components are and their respective relative weights.

    Your comment on trusting the "results" puzzles me a bit. I may be wrong, but it sounds like, if the stat confirms what my eyes see, it's a good stat even if I don't know how it works. If there's one thing I've learned in my time on message boards, it's that everyone sees thing through their own unique set of lenses. If everyone evaluates stats on how the numbers correspond to truth as they perceive it, it seems to me that stats then have the same value as opinions. Then again, maybe I misunderstood you.
     
  4. such sweet thunder

    such sweet thunder Member Staff Member Moderator

    Joined:
    Nov 8, 2007
    Messages:
    3,509
    Likes Received:
    78
    Trophy Points:
    48
    One stat topic I've been interested in and haven't seen addressed in a comprehensive way is the relationship between point differential and wins in close games. Hollinger has made point differential a heavily rated factor in his automated team rankings. I suspect that point differential is strongly predictive of a team's overall strength, but perhaps overweighted in Hollinger's stats to create false volitility. Hollinger is a journalist and it provides a good opportunity for stories.

    Regardless, if you watched the Heat last night, you were able to see how much they strugle, and have struggled in the past, closing out games. I think they're 1-8 in games decided by five points. Wade and Lebron struggled to figure out who should be the go to guy in late situations, especially since both of them are capable of accepting the role.

    So, losing close games is a team weakness, but it's a factor that really isn't given weight if your ranking is based on point differential.

    Does this make sense?
     
  5. JayJohnstone

    JayJohnstone Active Member

    Joined:
    Feb 15, 2008
    Messages:
    1,032
    Likes Received:
    3
    Trophy Points:
    38
    I think Hollinger would say that there is no such characteristic as "tendency to lose close games".

    I do think his reliance on point differential is overrating the Heat. When Heat are rolling, scrubs are going to get a lot of time and Heat has a ton of three point shooters. James Jones and House are going to put in a lot of points when in blowouts when teams stop playing D.
     
  6. JayJohnstone

    JayJohnstone Active Member

    Joined:
    Feb 15, 2008
    Messages:
    1,032
    Likes Received:
    3
    Trophy Points:
    38
    Transplant, sounds like a good article.
     
  7. transplant

    transplant Global Moderator Staff Member Global Moderator

    Joined:
    Jan 6, 2011
    Messages:
    4,111
    Likes Received:
    210
    Trophy Points:
    63
    I suspect Jay Johnstone's right about Hollinger's response, though I'm sure he could invent a stat for it if he put his mind to it.

    While I don't know of a single stat that does what you're looking for, a simple comparison of "Pythagorean Expected W-L" (basketball-reference.com just calls it "Expected W-L" now) to actual win-loss will at least get you in the neighborhood since the Expected W-L is purely point-differential-based. For example, both the Bulls and the Heat now have identical 31-14 records. The Bulls Expected W-L is 31-14 while the Heat's is 34-11. Tells me the Heat win-big/lose close (in relative terms).
     
  8. such sweet thunder

    such sweet thunder Member Staff Member Moderator

    Joined:
    Nov 8, 2007
    Messages:
    3,509
    Likes Received:
    78
    Trophy Points:
    48
    That seems improbable to me. If you do the math for a 1-8 record (assuming fifty percent chance of win or loss) there are only 10 possible outcomes, with 0 or 1 win, and 512 possible outcomes. That means, if you were to assume that there is no characteristic of losing close games, the Heat have fallen into the bottom 1.95% scenario. I think, at that point, you rely on your eyes and say there is a characteristic.
     
  9. such sweet thunder

    such sweet thunder Member Staff Member Moderator

    Joined:
    Nov 8, 2007
    Messages:
    3,509
    Likes Received:
    78
    Trophy Points:
    48
    And the more interesting point is does it matter? I would assume that this mild weakness of losing close games becomes a more serious flaw when the playoffs roll around and there are more tight games -- but that's all intuition.
     
  10. MikeDC

    MikeDC Member

    Joined:
    May 24, 2007
    Messages:
    5,643
    Likes Received:
    16
    Trophy Points:
    38
    Occupation:
    Professor
    Location:
    Indianapolis, IN
    Instead of results, I should have said something like "predictive power". For most any stat one develops, it should be possible to test it against lots of data. Take and stat you can think of, and test it against 10 years worth of box scores. How often does applyiing the stat pick the actual winner?
     
  11. MikeDC

    MikeDC Member

    Joined:
    May 24, 2007
    Messages:
    5,643
    Likes Received:
    16
    Trophy Points:
    38
    Occupation:
    Professor
    Location:
    Indianapolis, IN
    As a general point, I find most all of these sorts of stats interesting. The most misunderstand thing, I'd say, is not a particular stat itself, but the nature of most individual player statistics. To use a technical term, the results are endogenous, and people almost never understand this.

    What the hell does endogenous mean?

    It means the results we see are a product of their own interactions

    I recently saw this article comparing the scoring efficiency of high scoring players. It argues, basically, that a guy like Kevin Martin is way better than Carmelo Anthony or Derrick Rose, because Martin has higher TS% than his teammates, while Melo and Rose don't.

    I think this sort of analysis totally misses the boat.

    In this case, the "endogenous" results are measures of efficiency for each player.

    Being endogenous, the efficiency statistic of player 1 is somewhat the result of Playing with 2, 3, 4, and 5 and vice versa. And if 1 starts doing something differently, it’ll affect the quality of shots (and hence the efficiency) of 2-5.

    My objection is that Ziller’s advice (" So Rose could stand to pass more frequently. He needs to rein it in…") is premised on a change in the behavior of 1 not affecting player 2. That is, he’s not recognizing the endogeny of the variables.

    Let me give a concrete example. One of the 5 Bulls with a higher TS% than Rose is Ronnie Brewer. Lets suppose Thibs gets a clue and starts Ronnie, and he and Rose are on the court. Suppose Rose starts to drive to the basket, but sees Brewer wide open from 20 feet, and starts passing it over and over again, where before he (Derrick) would shoot.

    Pretty soon, you’ll start noticing changes in the efficiency statistics. Brewer will brick a bunch of those shots, and his TS% would go down. Rose, who’s not shooting it himself anymore unless it’s a real gimme, will see his TS% rise.

    This is also true even if we’re talking about good players used properly. Like, ok, maybe trading off between a Rose shot and Brewer taking long jumpers is sort of an extreme example, so let’s look at a closer example. Say Boozer and Rose.

    Should Rose be passing to Boozer more? Or what about Korver for a 3? Maybe, but it’s not so clear cut as Tom makes it out in the article I linked above. Teams watch tape and study hard on how to defend their opponents. And throughout games, they throw out different strategies to stop a guy like Rose. And at the margin, if they think he’s going to pass more and shoot less, they’ll employ fewer double teams and trap a bit less, and stay at home on the other Bulls a bit more. Which means they’ll have more contested shots and less open ones.

    While Ziller looks at Rose's stats as a sign of inefficiency and selfishness, I look at them and say Rose’s stats are a measure of his efficiency in the face of the dramatic inefficiency of his teammates. He’s out there for several minutes a game with Bogans, who at .480 TS% can’t hit the broad side of a barn. That number is especially inefficient because he’s almost entirely unguarded most of the time. And where is that guy who would be guarding Bogans if he were any kind of threat? Well, guarding Rose, and making him take harder shots.
     
  12. transplant

    transplant Global Moderator Staff Member Global Moderator

    Joined:
    Jan 6, 2011
    Messages:
    4,111
    Likes Received:
    210
    Trophy Points:
    63
    First off, great post, and though I'm 56 and have what I consider a fairly extensive vocabulary, you even taught me a new word. Thanks. Endogenous...cool.

    And I'm right with you.

    From much of what I've read lately, "scoring efficiency" is the new Holy Grail and TS% is Jason's magical sword. God knows, Rose critics have sliced and diced him with it repeatedly. However, few of those who so conspicuously kneel at the scoring efficiency altar point out that the NBA's TS% leader is Nene, with his teammate Aaron Afflalo close behind...while Melo is the unquestioned best player on that Nuggets team.

    Martin is clearly one of the darlings of the efficiency experts. As I see it, if you need only 6 points, rather than 10 points in the 4th quarter of a close game to win, you may want to go with Martin over Rose or Melo...but if you absolutely have to have those 10 points...

    Could have something to do with the fact that the Bulls and Nuggets are good teams and the Kings aren't.

    Of course, none of this means that efficiency stats are bad in and of themselves, but yeah, they can be endogenous. Cool word.
     
  13. such sweet thunder

    such sweet thunder Member Staff Member Moderator

    Joined:
    Nov 8, 2007
    Messages:
    3,509
    Likes Received:
    78
    Trophy Points:
    48
    I think Hollinger tries to correct for this with his usage statistic, presumably using some kind of curve. But I completely agree with you. The golden egg for coaching is how scalable a players numbers are: Who should get more or less shots and where should they come from? It's the question that separates good coaches from bad coaches, not a constant upon which you can make safe predictions. I have to imagine there's so much static that it makes most advanced statistics interesting, but of questionable merit when trying to make predictions.
     
    Last edited: Jan 28, 2011
  14. JayJohnstone

    JayJohnstone Active Member

    Joined:
    Feb 15, 2008
    Messages:
    1,032
    Likes Received:
    3
    Trophy Points:
    38
    I don't think we are on the same page.. Hollinger thinks W-L is noisy even adjusted for strength of schedule because a few fluke plays turns a W into a L or vice versa. Also, being able to beat an opponent by a large margin and not ever getting blown out are also indicators. He also thinks clutch ability is overated. He much prefers point differential to W-L because he feels it is less noisy.

    So if he couldn't get the data for point differential, I'm sure he would use W-L in his projects but doesn't need it if he has the point differential.
     
  15. JayJohnstone

    JayJohnstone Active Member

    Joined:
    Feb 15, 2008
    Messages:
    1,032
    Likes Received:
    3
    Trophy Points:
    38
    Yes, Hollinger is more enlightened than TS%. His PER combines Usuage plus stuff that would go into TS%.
     
  16. JayJohnstone

    JayJohnstone Active Member

    Joined:
    Feb 15, 2008
    Messages:
    1,032
    Likes Received:
    3
    Trophy Points:
    38
  17. Denny Crane

    Denny Crane It's not even loaded! Staff Member Administrator

    Joined:
    May 24, 2007
    Messages:
    72,957
    Likes Received:
    10,620
    Trophy Points:
    113
    Occupation:
    Never lost a case
    Location:
    Boston Legal
    Point differential is interesting, but not the whole story. A team with a +6 point differential is likely to be a top contending team.

    But what does it say when you have two teams:

    point differential 86-80 (average score) = 6
    point differential 106-100 (average score) = 6

    Or even a third team:
    point differential 110-98 (average score) = 12
     
  18. JayJohnstone

    JayJohnstone Active Member

    Joined:
    Feb 15, 2008
    Messages:
    1,032
    Likes Received:
    3
    Trophy Points:
    38
    Yea, that's why all the stat jockeys adjust for pace (possessions) when comparing things.
     
  19. Denny Crane

    Denny Crane It's not even loaded! Staff Member Administrator

    Joined:
    May 24, 2007
    Messages:
    72,957
    Likes Received:
    10,620
    Trophy Points:
    113
    Occupation:
    Never lost a case
    Location:
    Boston Legal
    My rule of thumb is you don't take any one stat in a vacuum. They all have some meaning, but the bigger picture is found in a combination of the stats, some old style and some new school.

    You do have to have some handle on what the stats mean. A lot of the stats on 82games.com have no real meaning. I once wrote the guys who run it and asked them what "hands rating" means and they don't have any real idea themselves. And they made up the formula :)
     
  20. transplant

    transplant Global Moderator Staff Member Global Moderator

    Joined:
    Jan 6, 2011
    Messages:
    4,111
    Likes Received:
    210
    Trophy Points:
    63
    Interesting discussion and please keep it up...if your friends and neighbors have thoughts, tell 'em to join in.

    And thanks to all who have participated so far.

    I'd encourage you to not simply reply to the preceding post but also go back to the OP. I'm looking for any and all opinions, thoughts and muses. I know it's a shameless request for material that I will without question steal (with or without attribution), but really, what are friends for?
     

Share This Page