March 8, 2011

Basketball stats and teaching stats

Here's an article by Dave Johns in Slate about NBA statistics. A huge amount of intellectual effort has been expended in recent years to bring basketball up to speed statistically with baseball. There has been a lot of progress, but the Holy Grail goal of coming up with a foolproof system for ranking players still has no consensus. 

For example, how good is Kevin Love (the 22-year-old Minnesota center who is an offshoot of the extended clan of Loves and Wilsons famous for the Beach Boys)? For example, Offensive Win Shares rates him as the best offensive player in the league and fifth best overall in total Win Shares.  Offensive Rating sees him as the fifth best offensive player in the league. Player Efficiency Rating says he's the third best player in the league. Other rating systems don't see him in the top ten.

Johns likes to disparage Love to show the problems being dealt with by the sophisticated statistics:
Rebounds also suffer from so-called "diminishing returns"—the idea that players on the same team effectively compete with one another for boards. Often a particular player—say, Minnesota Timberwolves center Kevin Love—serves as his team's designated glass-cleaner, and he scoops up balls that his teammates might well have grabbed anyway.

Okay, but, presumably, coach Kurt Rambis tells his players to let Love grab the easy rebounds (such as missed free throws by the other team) because he has such a good outlet pass, which ought to count for something in an overall ranking, right? Moreover, Love's offensive rebounding statistics are stellar, and there aren't all that many easy offensive rebounds.
The [plus-minus] technique can also examine the impact of top rebounders: Kevin Love consistently rebounds in double digits, but his contribution to his team's total boards is only about two to three per game, according to one analysis.

But the analysis Johns links to shows Love as being the best rebounder in the league by a margin of about 20% over the second best rebounder.

But, old fashioned stats can give a more well-rounded picture of Love than advanced rankings. This year, Love is on track to become the first player since Moses Malone's last MVP season in 1983 to average over 20 points and 15 rebounds per game (He's currently at 20.9 and 15.8). He's making 42.7% of three pointers and 86.2% of free throws, which are outstanding percentages for a center.

On the other hand, Love is under 6'8" in his bare feet and is a white guy who can't jump all that great: he doesn't block shots (only 0.4 per game, which is really low for somebody with so many rebounds).

(By the way, my impression is that rebounding correlates better with being a good all-around basketball player than does shot-blocking. The all-time bjg men like Russell, Chamberlain, Kareem, and Walton tended to be great at both rebounding and shot-blocking, but lots of guys are only good at one or the other. In general, the guys who are only good at shot-blocking are more often the weird Manute Bol-type talents. For example, on the playground, I was a pretty good shot blocker but I was an all-time awful rebounder. Partly it was getting pushed around by less skinny guys, but much of my rebounding deficit was cognitive: I never had the slightest clue where the ball was going to bounce. In contrast, I had a pretty good idea when somebody was going to shoot, so shot-blocking strikes me as pretty obvious while rebounding seems like a Dark Art. Your mileage may vary.)

Moreover, despite his superb hand-eye coordination, Love doesn't create much offensively down low (making only 48.3% of two pointers). His team, the Minnesota Timberwolves, has a very bad won-loss record, 15-50, and gives up a lot of points.

The more I look at it, the more I come around to middlebrow sportswriter Bill Simmons's convenient conclusion that basketball's old-fashioned box score stats are quite useful and Holy Grail one-number ranking statistics haven't yet gotten there. Looking at his non-advanced stats, Love's unusual combination of bulk and touch makes him look like a guy who could be extremely useful on a good team (like, say, Bill Laimbeer on the Detroit Pistons of the late 1980s) but who (at least not yet in his quickly evolving career) can't be expected to carry a bad team the way, say, Kobe Bryant carried an awful Lakers team with Smush Parker at point guard to a winning record a half decade ago.

On the other hand, there's a lot of learning available from advanced stats that don't try to rank everybody, but just try to look at elements of performance, such as a player's shooting percentage from the left or right sides of the court.

Much less intelligence has been devoted to analyzing teacher performance. Much of the recent work has been devoted to, yes, the Holy Grail of ranking teachers on Value Added so that bad teachers can be fired and good teachers rewarded.

The New York Times has an article on a hard-working NYC 7th grade English teacher at a prestigious public school. She has two Ivy League degrees and is much admired by her students, many of whom qualify for Stuyvesant. Yet, she only ranks at the 7th percentile among all NYC teachers. The article shows the complex formula used in the calculation, which appears to baffle most teachers. My guess would be that her students arrive so far above average that there's almost nowhere for them to go but down.

I'm always looking for sports analogies for social science statistics, since Americans think harder about sports. Being a teacher isn't really like being a player, it's more like being a coach. Probably the closest analogy is being a coach in a big high school with separate Freshman, Sophomore, and Varsity football teams. If you are the Sophomore team's coach, you more or less inherit the Freshman team's players (although the best will be sent up to the Varsity). So, if your Sophomore team consistently winds up with a worse record than the same players achieved on the Freshman team, your job will be in trouble.

For example, say the Freshman coach is great and routinely goes 9-0. All you can do as Sophomore coach is match his record or do worse, which might be an analogy for this New York teacher.

Similarly, the search for a super-sophisticated single number ranking system for teachers can overlook the advantages of less ambitious statistics at pointing out particular strengths and weaknesses, which would be of use both to teacher looking to improve and to administrators looking to maximize the usefulness of a teacher's talents.

19 comments:

  1. Just out of curiosity, do you happen to like any other basketball players who aren't white? In a field dominated by blacks, you seem to only know the white players.

    ReplyDelete
  2. The teacher in the NYT article should not worry about tenure - if she is so stellar as the article says, the article should be a perfect job application form for a private school.

    At least if there is some brains somewhere.

    Another thing: the formula, as written in the NYT article has problems besides its complexity. But that is another article for Steve.

    ReplyDelete
  3. Kevin Love was one of the most popular players in SoCal history. He's obviously of local interest.

    Bill Simmons scored a 1400 on his SATs and went to Holy Cross. It's a credit to him that he has written middlebrow and attracted such a large audience.

    Finally, and of note, Kevin Love has appeared on Simmons podcast (The BS Report) like 4 times. He's a great guest and Simmons and Love get a long extremely well.

    ReplyDelete
  4. This post prompted me to learn a (to me) astonishing fact about Bill Laimbeer: I thought I remembered that his father went to Harvard (which he did), and I checked the Wikipedia entry to see where the son had gone to school... Only to find this shocker: "Laimbeer played one of the three Sleestak characters in the Sid and Marty Krofft television show Land of the Lost after high school."

    Oh: and @Anonymous

    If you were really following this blog, you'd know that Mr. Sailer also knows about Manute Bol, who was the color of eggplant.

    ReplyDelete
  5. Your analogy with high school sports doesn't quite fit. Some schools tend send kids directly to varsity while some send them to JV or Fresh/Soph.

    So, the pool of kids that you are competing against is not stable. It's not as if the same kids are competing with the same kids at each level.

    ReplyDelete
  6. What I'd be interested in would be teacher statistics that could be used not to rank teachers from a scale of good to bad, but to assess teachers for fits to different types of classes: he'd be good for handling smart but difficult-to-manage kids, she's great in small groups with dedicated students, etc.

    ReplyDelete
  7. Love is a good player, in fact his stats will probably get him a max contract, or at least near; this is where the problems start.

    The agent for a guy averaging 20 points and 15 rebounds a game will demand a long-term 15+ million dollar deal. That dictates that as the highest paid player on the team, he will also, probably be the best.

    The problem with Love is that what he does does not lead a team to wins. 15 boards is a very impressive stats on it's face, but on further review, he plays in a turbocharged fast-break offense that allows more shots by both teams. He plays for the worst defensive team in the league which allows the other team many early shot-clock opportunities, and his own players are poor shooters, which lead to more offensive rebounding opportunities.

    Additionally Love is one of the worst defenders in the league, at a position (power forward, not center, Steve) that generally demands defense.

    BTW, neither blocked shots nor steals are considered the end-all of good defense; Rodman, for all of his leaping abillity never blocked shots.

    Those stats are are more of an indication of athleticism and tendancy to gamble. Guys who block a lot of shots are often considered good defenders but not always: Love's center, Darko, leads the league in blocked shots per game.

    ReplyDelete
  8. I don't think Basketball stats will ever be as comprehensive as those of Baseball. Nor will those of Football.

    Baseball is a unique sport that seems to have been invented by someone with a knowledge of experimental design. In basketball and football everybody moves at once. So when someone gets free and scores you never really know what happened. The scorer might have done well or perhaps it was the blockers. Or since both of these sports rely on deception, maybe the defense was fooled by the actions of someone else not in on the play.

    But baseball controls the confounding variables.

    In proto-baseball we can imagine that the team batted as a group. The opposing pitcher tossed the ball over the other team - like the throwing of the bouquet at a wedding - and all the batters swung at it. In such a game you would never know who was the best batter.

    You would also need a lot more batting helmets.

    Albertosaurus

    ReplyDelete
  9. Yeah, I saw the Sleestak thing, too at 4 am. One of the other Sleestaks was played by David Greenwood of Verbum Dei HS, who was Laimbeer's archrival big man in SoCal basketball and spent 12 years in the NBA.

    ReplyDelete
  10. In another vein - why all the concern with the quality of teachers?

    You have amassed evidence for years that most of the explained variance in the study of learning came from the student side of the equation not the teacher side. You would seem to be the last person on earth who would be panicked by low test scores into reflexively devoting more resources to the teaching side.

    When I was a Social Worker in the sixties in San Francisco I used to tell my fellow workers that I could solve the welfare problem. I suggested that we just replace all the City's Blacks with Chinese.

    VoilĂ . No more Welfare. The Chinese - even though they were the poorest group in the City - never went on Welfare.

    So in a similar spirit if you have a failing inner city school - just remove the Blacks and bring in Chinese replacements. Keep the same teachers and administration - they don't matter much. With a new student body the scores will be just fine. You - and most readers - know this to be true.

    Arthur Jensen predicted today's poor school scores in 1968. Yet somehow we have been encouraged to seek magic solutions to an imagined teacher shortfalls. Current teachers and current teacher remuneration is sufficient for teaching Whites and Asians.

    Albertosaurus

    ReplyDelete
  11. "I suggested that we just replace all the City's Blacks with Chinese..."

    Brilliant idea!

    And if we replaced California's whites with Chinese, we'd do the same with disabillity.

    ReplyDelete
  12. "Just out of curiosity, do you happen to like any other basketball players who aren't white? In a field dominated by blacks, you seem to only know the white players."

    If a young relative of Michael and the other Jacksons was setting NASCAR per lap speed records but not winning many races, that would be pretty interesting too.

    ReplyDelete
  13. BigBabyJesus3/8/11, 12:54 PM

    Love is a power forward, Steve.

    "The problem with Love is that what he does does not lead a team to wins. 15 boards is a very impressive stats on it's face, but on further review, he plays in a turbocharged fast-break offense that allows more shots by both teams. He plays for the worst defensive team in the league which allows the other team many early shot-clock opportunities, and his own players are poor shooters, which lead to more offensive rebounding opportunities. "

    I'm pretty sure Love leads the league in rebounding rate, the percentage of available rebounds, too. His rebounding stats are not a mirage, but it is almost certainly true that he's not a very good defender.

    ReplyDelete
  14. "Another thing: the formula, as written in the NYT article has problems besides its complexity. But that is another article for Steve."

    Any help would be appreciated.

    ReplyDelete
  15. "And if we replaced California's whites with Chinese, we'd do the same with disabillity."

    Even the Jews?

    Oh right, you don't like having your stupidity pointed out.

    ReplyDelete
  16. Basketball really isn't a "Stat" game, cause you can't quantify defense or nuance.

    Joe Blow brings the ball up court, passes the ball to X who makes an incredible falling backwards, over the head 20 footer that barely touches the net. OTOH, Magic throws a behind the back, thread the needle pass to Kurt Rambis who makes a layup.

    Both are graded as assists.

    Other than offensive rebounds, 3 point FG percentage and FT percentage most NBA stats are worthless.

    I loved Dennis Johnson of the 80s Celtics. Stats told you nothing of what was good or bad about his game. He was a great Defensive player - no stats on that. He brought the ball up, sometimes under full court press and ran the offense - no stats on that. And he didn't get many assists because Bird and McHale were great passers but he was a great passer.

    OTOH, he had a good FG percentage, which was misleading. No one missed more open jump shots than DJ. Sometimes he'd get hot and the Celtics would walk away, but the bottom line is when you throw in the Fast-break layups and open jumpers DJ's STATS completely misrepresented what a hot/cold shooter he was.

    ReplyDelete
  17. I like the topic.

    Are there any American born whites in the NBA who are as good as Nash, Nowitzki, or Ginobili?

    Goatweed

    ReplyDelete
  18. "Even the Jews?"

    Are Jews "whites" this week?


    I lost the WN calendar Albertosaurus sent me.

    ReplyDelete
  19. "Are there any American born whites in the NBA who are as good as Nash, Nowitzki, or Ginobili?"

    Or Pau Gasol ...

    Yeah, that's an interesting question. Why are tall white guys raised in all white environments having better NBA careers than tall white guys raised in environments where blacks dominate high-level youth basketball? So, Love is an interesting exception to thir new rule (of course he grew up in the most famous liberal whitopia of Portland).

    ReplyDelete

Comments are moderated, at whim.