-VS-
Game Time Start: 8:00 PM ET
Where: Chesapeake Energy Arena, Oklahoma City, OK
Officials: J. Goble, K. Scott, S. Wall
Media Notes: Indiana Notes, Oklahoma Notes
Television: FOX Sports Indiana / FOX Sports Oklahoma
Radio: WFNI 1070 AM, 107.5 FM / WWLS 98.1 FM
NBA Feeds:
*NBA Audio League Pass (available free to NBA All-Access members)
*NBA League Pass Broadband (subscription req'd)
*NBA League Pass Broadband (subscription req'd)
REMINDER: Per PD policy, please do not share a link to, describe how to search for, request a link to, or request a PM about streaming video of a NBA game that is not coming directly through the NBA. Not even in a "wink-wink, nudge-nudge, know-what-I-mean" round-about sort of way. Thank you
|
|
PACERS Paul George - Fractured Right Tibia/Fibula (out) THUNDER Steven Adams - right hand fracture (out) Kevin Durant - foot surgery (out) |
Seth Partnow: Demystification of DIY — Defining Basketball Analytics Down The furor over analytics (re-)sparked by Charles Barkley’s pre-All-Star Week tirade has gotten us here at Nylon Calculus talking. Moving past the immediate, defensive reactions to Barkley’s particular perspective, there was something to be taken from the discussion. Barkley and many others don’t really know what “analytics” entail. Partially, that’s an error on their part, being paid to observe and comment on the league as they are. But it’s also on “us,” the numbers-driven community. While things like points, assists and rebounds are analytic tools, they are digestible, non-threatening, and perhaps most importantly, easily related to basketball as viewed – LeBron James scored two points on that dunk, sweet dime by Ricky Rubio, Andre Drummond had twenty rebounds again. There’s a reason for the dominance of box score statistics that goes beyond sheer repetition and tradition. They are basketball. On the other end of the spectrum are highly involved math-and-formal-statistics intensive systems. These produce things like ESPN’s Real Plus Minus, or Andrew’s Player Tracking Plus Minus, and form the basis of many of the research papers forthcoming at Sloan. These methodologies have their uses. But they can be technically daunting, prone to misapplication and interpretation, and exceptionally difficult to differentiate in lay terms. Hell, the sentence explaining their limitations is daunting. And it’s a key point; how is a basketball lifer with a different approach to the game, or even a more willing student, to decide which set of black-box numbers to believe? That there is perceived to be a wall between the two camps, or between numbers-intensive analysis and game film-based scouting, then, is largely a failure on the side of analytics. Contrary to the beliefs of some voices hollering in the wilderness, the professionals were doing things pretty ok before the quants came along. There were certainly areas for improvement, but they weren’t just throwing darts. Most of the players believed to be great under the new school were recognized as such by the old school and vice versa. Traditional scouting has done a reasonably good job, on aggregate, of slotting players into the right order in the draft.5 The new methods must be demonstrated to be an improvement. The onus is on the new methods to explain themselves. In the legal arena this is known as the burden of persuasion. Persuasion. Yet the language used is often more exclusionary than inviting. “Metrics,” “studies,” and “models” are fancy sounding words that serve to simultaneously make the achievements seem more impressive but also less welcoming. The use of this terminology is understandable: describing something as a “metric” rather than just a “stat” is meant to imply a certain progressiveness of thought, a signifier that I’ve moved beyond “Yay, points!” as a good means of judging talent. To some degree, it’s perhaps a more accurate use of the language. But we probably go too far too often and end up sounding more than a little douchey to the not-already-converted. Perception is a two-way street, after all. It’s also important to be clear on what analytics is not. It’s refinement, not reinvention. For as much as Barkley and other commentators bemoan the lack of a midrange game, I think they would agree with this fundamental notion: teams should get good shots. Layups and dunks are good shots. Wide open three pointers are good shots. We don’t need MoreyBall to tell us this. Moreover, the “insight” that a midrange shot is lower efficiency than a dunk tells us virtually nothing about how to produce more dunks from an offense. When Hall of Famers scoff at the notion that the midrange is somehow less than ideal, they’re not scoffing at the idea that teams should take good shots. They’re amused and probably a little pissed off at the idea that getting to the rim is somehow easy. Efficiency is difficult to come by; no one knows that better than a 6’4″ power forward who made a career of fighting for rebounds — and, somewhat incongruously, taking a lot of 3s. Further, while some of the insights coming forth are truly PhD-level, most of it isn’t that hard. At least, not that hard from a technical or mathematical perspective. It’s much more about the logic. “What question am I trying to answer?” is often the most important question, followed closely by, “Do the tools I’ve chosen answer that as well as possible, given what’s available?” Or to put it more simply, the difficulty isn’t in the math. It’s the basketball, silly. If a stat/metric/number/analytic/gizmo can’t be related back to basketball in a recognizable way, there is no good reason for someone already knowledgeable of the game to accept it. To paraphrase an old saw, there are lies, damned lies and bad analytics. Even a properly specified, rigorously tested and logically sound model that we as analytics types know is meaningful is a lot to digest from a cold start. An analogy I’ve used before is one does not learn the guitar by playing Jimi Hendrix licks, rather working their way up to that point. And that ramp up is maybe where the “community” has lost its way a bit. Our founding mission here includes offering up numbers-based work from the perspective of both the rocket scientist and the raconteur. The deep-dives and the quick hits. As is perhaps too often the case in the community, we’ve erred a little on the side of the abstruse. [Ed. note: Mostly by using words like abstruse, SETH.] Part of allowing a more inviting point of entry into a quantitative mode of enjoying the game,8 is demystifying this stuff. For example, almost nothing I publish here involves anything more conceptually difficult from a numbers standpoint than compound multiplication and division. Partially, this is because I’m more interested in discrete areas of study like “who’s getting the most open shots?” or “which players have been most effective defending the paint this season?” — the kind of thing you’d talk about over beers with a friend. For the purposes of fan-serving analytics, good enough really is good enough. Most analysis should not be taken as a rank ordering, but rather more of a tiering system. While Rudy Gobert has been at the top of the Rim Protection charts by my methodology over the course of the season, I don’t think it’s a demonstrated fact he’s “better” than Roy Hibbert. He may have performed better in certain ways over the course of the season10, but both have been very good and any direct comparison is best left there. It’s an application of the 80-20 rule, in that a good first approximation of the answer being sought is often available reasonably quickly, but that last fine tuned bit of accuracy is going to take a disproportionate amount of work. But, again, that 80% answer is often a pretty good one. And that’s true here. This post isn’t an attack against anyone; if anything, it’s the extension of an olive branch. We all love basketball; if there’s been confrontation, it’s in the past. But if you want to dive a little deeper and take a peek behind our curtain, keep reading, because there’s some fun stuff ahead. Even internally here at Nylon Calculus, we often mistakenly see wizardry where there has simply been thoughtful application of tools and questions. In examining how the addition of a third ball-hungry guard had disrupted Phoenix from last season to this, Jacob was momentarily taken aback by a stat I’ve been using for Time of Possession %, or a different way of looking at how much different guys have had the ball in their hands on offense. He wanted to know how I calculated it and was surprised to learn it was simply SportVU’s time of possession divided by minutes played to allow for a meaningful comparison across players with different minute loads. A simple calculation for a simple, descriptive stat which allows a rough measurement of how often various players possess the ball. Complexity and usefulness don’t always go hand in hand...CONTINUE READING NYLON CALCULUS |
Pacers Candace Buckner @CandaceDBuckner Jared Wade @8pts9secs Tim Donahue @TimDonahue8p9s Tom Lewis @indycornrows Ian Levy @HickoryHigh Whitney @its_whitney |
Thunder Darnell Mayberry @DarnellMayberry Royce Young @royceyoung J.A. Sherman @WTLC Zebulun Benbrook @ZebulunBenbrook Tyler Parker @_tkparker Trey Hunter @TreyHunter87 |
Comment