The Houston Rockets and the methodological dangers of Reductionism

I am currently watching a Tarantino-style murder scene: at the half of Game 1 of the 2016 NBA Western Conference Quarterfinals, the Houston Rockets are down 33-60 to Golden State. No team has scored fewer points in the first half against this Warriors team, and 27 points it the Dubs’ highest lead at the half this season. To put it bluntly, Houston has been abysmal: James Harden is 2-9, Beverley 1-6, Brewer 0-4, with fourteen first-half turnovers and a putrid 16.7% shooting from beyond the arc. Dwight Howard’s back is shot, despite his best “it’s a contract year” efforts, and Michael Beasley and a washed-up Josh Smith are their best bench options. More than that, no one on the team appears to even be trying: this team limped into the playoffs on a 41-41 record, and has duly collapsed competitively in the face of the Warriors onslaught.

It wasn’t supposed to be this way, and most of us didn’t expect it to be: seasoned NBA commentators, from Matt Moore at CBS to Bill Simmons called the Rockets to improve on their appearance in last season’s Conference Finals and pose one of the greater challenges to the Warriors’ supremacy this time around – the team stole Ty Lawson in a trade from Denver, James Harden was coming off a career year as runner-up for MVP, and they’d run Golden State close in the first two games of the Conference Finals before mentally collapsing in the face of the perimeter-oriented barrage. Here’s what Armchair’s own Serge Leshchuk wrote in his predictions for the season:

With Lawson, they have a secondary playmaker who can take the pressure off Harden and force opposing guards to defend. When healthy, Howard is a top five center in the league (the other point is that he’s hardly ever healthy). James Harden is an MVP contender. Add all of that up with the fact that they have some of the best role-players built specifically around their superstars and they are as good as anyone in the West.

Instead, the whole project soured rapidly: Harden spent the season looking utterly disinterested on defense, Ty Lawson proved an ineffective tire-fire that was promptly waived at the deadline, and their role players struggled all season to hit shots. Kevin McHale was fired after a 4-7 start to the season, replaced by former assistant J.B. Bickerstaff for the remainder of the season. Dwight Howard is unlikely to re-sign with Houston, having become a shadow of his former self since his back went, and the palpable sense from watching Rockets games is that this team just doesn’t like playing with each other. Everything is bad: the decision-making, the shots, the communication (I have watched multiple players scramble for the rebound, get in each other’s way, and commit the turnover about five times in this half), the movement, and the execution on shots at all areas of the court. This whole team has the feel of being on the verge of being blown up; nobody cares, nobody tries, and nothing has quite gone according to plan.

The broad reasons for this collapse aren’t that hard to identify; we’ve talked about them before, multiple times. Daryl Morey is a disciple of advanced metrics, having preached them as a means of making the Rockets competitive without having to tank for good draft picks. In theory, the process works a bit like fantasy basketball: stockpile players with a high upside, wait for them to get hot, ship them for better players, repeat incrementally until you end up turning Kevin Martin, Jeremy Lamb, and a couple of picks into James Harden in 2012. The system works equally well – on paper – for role players: find players who shoot good percentages, and surround your stars with them to produce a team that can get hot in a hurry. In essence, play the percentages: take open 3s, protect the rim, and get to the line. The system also seemed to work last year: it was Josh Smith, acquired for pennies on the dollar via waivers from the Detroit Pistons midway through the year, who ignited the Rockets in Game 6 against the Clippers to claw back into the game and the series; it was Trevor Ariza and Corey Brewer who sank the Mavericks in the first round from deep, while Dwight Howard bullied Tyson Chandler on the glass for the series.

The problem for the Rockets has not been an analytical one: rather, it has been systemic failures that are difficult to quantify that have sunk their season. In March, Bickerstaff called them “a broken, fragmented group.” Simply put, this group of players don’t like playing with each other: James Harden’s astronomically high usage rate – third-highest in the NBA – is great for getting points at the free throw line, but it also deflates team chemistry when nobody else ever seems to get touches for long stretches of the game. Dwight Howard is a great statistical rim protector, but defending elite teams – particularly ones like Golden State and San Antonio that can move the ball quickly –requires communication and teamwork, whether that’s switching on pick and rolls or doubling players or covering the open corner shooter, and those things don’t happen when players don’t enjoy communicating with one another. That becomes difficult when players clearly don’t like playing together, and when dressing room personalities clearly are out of sync. Ty Lawson was a great addition on paper, with 9.6 assists per game the previous season, but his personality and well-documented off-court issues proved poison to the Rockets’ chemistry. For role players, that Morey views them as assets to be traded and speculated on deflates that chemistry even further: players who don’t seem themselves as important to a team long-term often stop trying. Watching Corey Brewer and Josh Smith play defense right now suggests that this is the case here.

But even beyond basketball, I think the Rockets’ struggles this season are symptomatic of the broader failures of the reductionist methodology – the notion that one can determine the outcome of a whole by looking at individual components and aggregating them. Sometimes, having a number of good individual shooters and defenders doesn’t work; at a tactical level, part of the difference between the 2014 Spurs and the 2016 Rockets, both of whom ostensibly used a motion-heavy, percentages-based offense with lots of 3&D guys is that San Antonio’s players like playing together, and the personalities mesh well, and having the greatest power forward of all time on your team has effects that go beyond the stat-sheet. Until we figure out a way to quantify ‘chemistry’ in a manner that’s more complex than a 2k stat, that will be a factor that is impossible to predict based purely on the largely analytical model that Houston has relied upon.

This kind of analytical oversight, where the aggregating of individual stats fails to account for systemic factors like locker room personality, is something that comes up a lot in my ‘home’ field of International Political Economy: Open Economy Politics, the dominant theoretical school of IPE on this side of the Atlantic, makes a similar reductionist gamble, focusing on the ways in which the interests and preferences of domestic political groups (for example, dairy farmers in Quebec and Parma) aggregate up to influence the structure of international politics (i.e. World Trade Organization rules around agricultural protectionism). But in doing so, OEP often misses out on what are termed ‘endogenous factors’ – elements of international politics that can’t be measured by breaking them down into their constituent, national parts. In the IPE of trade, that often means the way in which ideas about ‘good’ economics or ‘good’ politics shape what sort of deals countries are willing to make with each other, or the ways in which the notion of an entity like the EU ‘winning’ or ‘losing’ an agreement like CETA impacts how the EU negotiates TTIP with the United States. In the case of the Rockets, aggregating up shooting percentages and defensive statistics failed to account for the way these players would play together, that team defense means trusting your teammates and liking them.

But there’s another problem with the analytics-driven view of player quality, insofar as it offers a snapshot of a player’s profile. This is similar to the problem that I’ve encountered in trying to use Liberal Intergovernmentalism (LI) – a theory of economic and political integration – to explain the European Stability Mechanism Treaty for my M.A. Thesis: LI views negotiations in a vacuum; states make bargaining decisions on the basis of immediate, economic factors, and delegate decision-making power to EU institutions based on whether doing so fulfills those material interests. However, such a ‘snapshot’ analysis often misses out on factors that accumulate over time – most critically for my studies, the question of whether the outcomes of previous treaty negotiations shape how states go into bargaining on subsequent negotiations. Similarly, letting analytics be the modus operandi of your front office ignores the noise that is often present in statistical measures: players have good shooting seasons – they get hot, or they play in a system that maximizes their efficiency, or they play with teammates who are good at recognizing a shooting streak and getting them the ball. All of those things get lost when you rely on the snapshot of analytics; Houston’s troubles could be a case of players like Trevor Ariza regressing to the mean, and the system of analytics simply not spotting that.

It’s an interesting constant in the dominant school of statistics-driven analytical thought in North America, be that in political economy or sports. I suppose the one big difference is the scale of the consequences: when OEP or LI analysis ignores systemic factors, it gets criticized in two or three academic journals, everyone involved adds a line to the ‘published works’ section of their CV, and life goes on. When Daryl Morey does it, the cornerstone of his professional life implodes on national television.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s