September is a funny time of year for basketball geeks whose interest in other sports ranges between “casual” and “yawn.” For true basketheads, all you can really do in September is a) watch EuroBasket and b) rehash old arguments.
We’ve all heard the same points laid out. We are accustomed, for example, to the angry fist-waving of the Fire Ty club and we have all but memorized the counters of the coach’s defenders. Both sides will keep at it indefinitely, but neither side has anything really new to say.
Several times in recent weeks, I’ve found myself pining for a better analytic system to help me in a discussion about some of the topic that are most difficult to quantify. I legitimately want to be better equipped with objective, empirical information that so I can either show that I’m right or know that I’m wrong on these vague questions. We’re approaching an era in basketball where tracking cameras can show us how each and every play is executed, and yet we still can’t, as a basketball community, answer some of the most fundamental questions about the game.
It’s the era of big data and we still have big questions that need answering.
Here are five areas where I hope some creative things are happening to answer burning questions and either validate me or shut me right up.
1. The value of coaching
How do we know which coaches are helping their teams win more games and which one aren’t? How do we know that the value of coaching isn’t some farcical concept altogether in this players’ league? Some hoop pundits think that, in any given year, there are 2-4 Popovichesque geniuses and the rest of the coaches in the league are just trying not to screw up. But how do we know?
I’ve heard a lot of ideas about how to measure the impact of coaches, but all of them stink in their own ways (the ideas, not the coaches). You can measure player/team performance against projected levels, but the problem there is that you might actually be measuring the quality of projections and then sticking the blame/credit with the coach. Or if a player just plain doesn’t show up ready to play, that reflects badly on the coach who may have had nothing to do with the injury, ego, drug problem or reality show that actually caused the player to fall short of expectations.
I’ve read suggestions that you could measure X-and-O prowess by looking at points per possession out of timeouts. That might only measure players’ adherence to the whiteboard, though, or undervalue coaches who have created such a strong culture they don’t need to timeout to right the ship. Also, late-game situations have lower efficiency overall, so that type of system might actually punish coaches who smartly saved their timeouts for use in tight situations late in games. Meanwhile, an idiot coach who burned through his timeouts earlier in non-pressure situations with higher efficiency averages would look like a tactical genius by comparison.
Really what I want to see is a coaching win share. How many of the Bulls’ 72 wins in that historic season can we credit to Phil Jackson? Did Ty Corbin cost the Jazz wins in 2012-13 or did he actually bolster their record? If someone can figure that out, lunch is on me.
2. Individual opponent strength
Right now most fans are extremely bullish on four youngsters who have played mostly bench roles against bench talent. Gordon Hayward and Derrick Favors have certainly seen some time against first-tier talent. For Alec Burks and Enes Kanter, a decisive majority of their minutes came against second-tier dudes.
But how much does that matter? We really don’t know because we don’t have a way to standardize that and compare Favors’ opponent load to, say, Al Jefferson’s. We can use opponent PER, but that can be spun both ways: is Kanter’s opponent PER so freakishly low because he’s that much better a defender than Jefferson, or because he was working against crappy centers while Jefferson dealt almost exclusively with starters? Probably a bit of both, but it would be easier to project Kanter’s season and future if we knew how much was due to either factor.
Somehow, we have to be able to mine existing data in way that we know that a guy spent X minutes guarding that guy, whose PER for the season was Y, and factor all that data in so we know if a guy looked better than he actually was. Otherwise, we might be comparing Mo Williams’ play against CP3 to Burks’ play versus Royal Ivey.
3. Impact on teammates
I’ve heard David Locke say that he believes there are three special guys each year who make their teammates better. (His guys last year were LeBron, CP3 and Durant.) I personally think that list is probably too short and that a multitude of players have a more subtle impact on their teammates’ ability to perform. But how do we know?
We may actually be closer to the destination on this. We can already track how Burks performs when Hayward is on or off the floor. All somebody needs to do – and someone will, soon – is figure out how to aggregate all those data points into some kind of an index on which players have a net positive impact on the play of others around them. This would be wildly interesting, both for short-term applications (who should be playing with whom to make their current team better) and long-term (which players should teams target).
4. Defense… period
It is ridiculous how rudimentary our defensive analytics tools are for this day and age. People have started to ask the right questions, but I don’t think we’re anywhere near the answer. Defensive Win Shares is a horribly simplistic stat that basically rewards minutes played and rebounds. Synergy is very interesting, but is only as accurate as the play definitions are clear. And, the guy logging the play and assigning a “primary defender” has to understand 30 teams’ defensive philosophies enough to know exactly which guy screwed up or deserves credit for the stop – highly unlikely.
Which is probably why the best answer to how to measure defense is to quantify some sort of behavioral system like what the Jazz were doing in the tail end of the Sloan days. On each play, they basically established on a Y/N basis for each player who did their defensive job within the collective defensive scheme. However, getting 30 teams to open their defensive playbooks so we can do the same thing at a league level is probably a pipe dream, so we’ll probably be stuck with an interim solution until then.
5. Defense by zones
In the last decade, offensive zones have caught fire. NBA.com now features shot charts and several other tools (Synergy, Hoopdata, etc.) measure what happens to a player’s offensive game at different distances and angles from the basket. But what about defense?
If a guy has a hard time on offense with a left-to-right motion, won’t he also struggle making that same motion as he mirrors his man’s right-to-left drive? What if certain players are better at closing off the right baseline than they are the left? Maybe Trey Burke will have an easier time fighting over a screen at the right elbow than he will crossing the paint on the left block. If any of those are the case, being able to better quantify that would not just help me win or lose arguments; it could change the way teams scout or even make decisions about when to help and in which situations you can trust a guy to close out his man without a second defender coming.
What’s on your analytics wish list? Do you have ideas on how we can start to solve for these five areas?