The introduction of analytics to Basketball
The revolution brought about by the use of data analytics in sports, particularly US sports is fairly well known. In 2011 a feature film, Moneyball, was made about Billy Beane, the then general manager of the Oakland Athletics. It told the story of his pioneering embrace of an analytics driven approach in order to ensure the smaller club he was running could compete against teams that just outspent them. They used analytics to identify players that traditional stats didn’t clearly show the value of. Those players were regularly far easier and cheaper to sign or trade for. When assembled together they performed at a level equal to teams with much larger payrolls and rosters of flasher ‘star’ players.
The NBA, one of the most progressive and trend setting sports leagues in the world, quickly took note and started using analytics to overcome the statistical challenges it had long faced. As a free flowing but also possession structured sport Basketball very easily lends itself to summary by discrete measurement. From the NBA’s very first season in 1946-47 specific player actions in each play were being recorded. By the 80’s every game could be summarised by its ‘box score’, which totalled each player’s contributions in a number of key ways. By the mid to late 2000’s insights from analytics began to really show the shortcomings of these simplistic summary statistics. For example, once the pace at which teams played could be measured it became clear that just because one player scored a few more points per game than another team’s player it didn’t mean they were a more effective scorer. Their team played at a faster pace resulting in more possessions, more opportunities to shoot and therefore more points scored. A faster pace however did not necessarily mean the scoring was more effective. As per minute and per possession advanced statistics grew in prominence increasingly long held opinions were shown to be faulty. At Butterfly data we’ve often found the new insights we can provide clients into their data can easily shake up the status quo.
In the 2013-14 NBA season technology was introduced that allowed the NBA to track the exact position of the players and the ball at every second of the game. This of course provided a massive amount of raw data that could be fed into analytical models. A key area it was able to inform was the analysis of defence. Rather than relying on the few traditional statistics (which tended to reward players taking risky gambles) models could now measure how effectively a defender forced an opponent into performing poorly. Despite its accuracy that kind of analysis has yet to gain a lot of traction in more mainstream Basketball analysis, likely due to the complexity of the concept compared to something as tangible as ‘stole the ball’. That quite likely reflects a more macro issue with data analytics, it often can only be as useful as its intended audience will allow it to be. On every project Butterfly Data undertakes one of our key objectives is ensuring end line users understand the tools we’ve built for them and can use them powerfully.
The Houston Rockets are the NBA team who, more than any other team, are driving their Basketball decisions by analytics. They’ve prioritised specific shots, dispensed with traditional player positions and installed an isolation heavy offense. In Basketball circles it’s been a highly controversial approach and despite some success has yet to fully prove itself. Critics argue that not only is it an unattractive style of play to watch, it ignores much of what traditional human insight and experience say. We passionately believe in the power and potential of data science at Butterfly Data but equally understand what got us to this place is valuable as well. Our on-site consultants work with our clients to ensure the analytics we provide build on their existing successes.
