How data and analytics is shaping the NBA
How data and and analytics is shaping the NBA
At the highest level of professional sport in the US, in the corridors of an NBA team’s headquarters, every decision is a billion dollar decision. According to Forbes, franchises such as the New York Knicks and Los Angeles Lakers are worth over $4 billion. Without on-court success, those figures mean less and less each year, as their millions of fans turn off their TVs or put their jerseys away. Teams are desperate to win, and with 29 other teams fighting for the same goal from the same global pool of players, teams are required to dig deep and find the slightest of winning edges.
Data analytics has been a prominent field in which to find those edges for every successful team in the 21st century. The thinking is to reduce the weight given to “gut feel” instinct on how good a player is, to figure out where inefficiencies and errors may lie in that evaluation, and thus use analytics to sort the undervalued from the rest. In professional sport, you only have so many players you can put in your jersey, and only so much money to spend on their salaries. Analytics offers a chance to make the wisest decisions under these constraints. Here are some of the approaches that NBA teams take with their data:
Predicting successful draft prospects
Every year, between seasons, NBA teams take turns to select from a pool of college and young international players in the NBA Draft. Unless they have traded away picks, each team gets two picks to add to their team. Hitting on these picks by selecting the right players is a crucial part of building depth, long-term stability, and championship potential in their franchise.
Historically, teams without a top three pick have really struggled to select accurately. There are countless ways to measure this struggle, but put it this way: USA Today found that from 1990-2014, 80 percent of top three picks averaged at least 10 points per game in the NBA, which means they were at least contributors. But from picks 8-11, only 42 percent of players averaged that 10 point mark, and from picks 12-15, just a third of players got there. What that means, is that as you go down the draft board, the chance of selecting a solid player soon becomes much less than a coin flip.
To increase their odds, teams turn to data. In a move that would have been unheard of in the early 2000s, teams now employ rooms full of people to run vast amounts of player physical and statistical data (height, weight, BMI, speed, wingspan, college stats, and more) into complicated models that predict a player’s success based on a database of thousands of past players’ profiles. Strong and detailed data analytics on these inputs gives the team an insight into which aspects of a prospect are predictive of success.
Although the teams’ models are proprietary and well-guarded, public models such as FiveThirtyEight’s CARMELO projection model show similar processes that teams would discuss in-house. Without being perfect, or the be-all and end-all, these models contribute to a collaborative process with more traditional qualitative scouting methods, e.g having experienced scouts watch the player and take notes on what they see. As analytical pioneer and General Manager of the Houston Rockets, Daryl Morey, says it: his team’s model isn’t necessarily the “right answer” but it’s a “better answer” than guesswork and instinct riddled with human error and bias. Morey has used an aggressive data-driven approach to optimise the Rockets’ drafting since 2007. Even in his first season, it paid off. Morey had the 26th and 31st picks in the 2007 NBA Draft, each of which holding a likelihood of drafting a starter at around 1% based on past drafts. With an unprecedented reliance on data, the Rockets selected Aaron Brooks and Carl Landry, both of whom became starters in the NBA.
Spatial player tracking
Morey was one of the first to implement such quantitative models in the early 2000’s, but as of 2020, the depth and complexity of the data has grown exponentially. Teams now utilise player tracking data to evaluate players, taking coordinates via video equipment and AI programs that track each player’s movements and actions throughout each game. With that tool at hand, teams can extract and understand much more context around each shot, pass, or rebound.
Much of this modern data is made public via the NBA’s official statistics website, such as by showing players’ shooting percentages at different distances from the hoop. Even deeper, the tracking data can show you that while Player A and Player B each average 10 rebounds per game, Player A usually grabs rebounds in plenty of space, while Player B consistently grabs rebounds despite being surrounded by opponents. By quantifying and analysing that data, teams can innovatively distinguish between players that they might draft or recruit in the offseason.
Three-point shooting revolution
As a result of statistical models and player tracking software, NBA teams have begun to adjust strategy both for recruitment and on-court tactics. One of those adjustments is to shoot more three-pointers, which are statistically proven to be the most efficient shot in basketball apart from dunks and layups, and much more efficient than the popular long two-pointer of earlier eras.
One of the innovations that Daryl Morey brought to basketball is the Houston Rockets’ historically high volume of three-point shots and reduction in long two-point shots, dubbed “Moreyball” after the mastermind behind it. The Rockets lean heavily into the data, and in 2018 they were the first team in NBA history to shoot more threes than twos in a season. Other teams are following suit: the number of threes shot by all teams has increased year-over-year since 2012.
With this evidence-based approach to team strategy in the modern game, teams are playing better offense than ever before, with offensive efficiency at record highs. As top NBA shooter Stephen Curry told The Ringer: three-point shooting is “infiltrating the league, younger generations and how they approach the game”. In many ways, it's a data-driven revolution that is here to stay.
Published by Will Eddowes