NBA Hackathon

Hey,

Edit: I’m well aware of the typos here. This was written pretty quickly and the character constraints were pretty high, which resulted in some fragments and otherwise weak grammatical choices. I apologize.

The NBA will be hosting a hackathon this Fall, and I’m applying to be a part of it! One of the questions on their application is “ Suppose you are the general manager of an expansion NBA franchise to debut in the 2016–2017 season. As part of the expansion process, the NBA conducts an expansion draft, from which the expansion franchise can select players from other NBA teams for its roster. What is the process you plan on using to (analytically) select or value players from other teams?* Which datasets, factors, variables, trends, and other considerations (e.g. salary cap) will you take into account? Which technological platform would be the best to utilize your strategy and why? Please be as specific as possible in your methodology description.”

I actually really liked this question so I wrote up a response, which you can read as follows:

In order to succeed in such a draft, I would need to ensure that I draft a team of individuals whose talents and attributes can be used to adequately contribute to create a team that is larger than the sum of its parts. My criteria for player selection would be position dependent, and for the purposes of this note, I will consider the five traditional positions, and make considerations about the trend towards positionless basketball afterwards. Starting with the point guard, I would look for an individual with strong linear speed and change of direction as per data collected at the NBA Combine. These stats would be used as a baseline to understand the defense such a player could provide on the perimeter, as well as their cutting ability. In addition, for more experienced guards, I would consider data collected by PlayerVU such as unassisted Field Goal% and Field Goal% of those receiving a pass from this player to determine how well the player can create for themselves and for their teammates. I would also evaluate point guards on their ability to draw up plays and analyze defenses in the film room. For a shooting guard, I would consider catch and shoot%, total distance run during an NBA game, average amount of time with the ball in hand, and field goal% within 5 feet of the rim as per PlayerVU. This data would help me evaluate the ability of the shooting guard to cut and move within an offense, while also allowing me to evaluate if they have tendencies to over dribble when they are with possession. In terms of a small forward, I would consider similar traits as I did for shooting guard, but I would also include more defensive-oriented stats, such as opposing 3FG%, and number of times opposing team shot the ball when SF was the primary defender. Because the league has many dominant forwards, it’s important to have an effective defender at the SF/PF position. For Power Forward, mobility is less of an expectation, although it is a benefit. Instead, I would consider Post FG%, 3FG%, and team FG% with PF as primary screener. Because much of the NBA is dominated by Pick and X play with a trend towards the PF serving as the screener, rather than the center, as has been the case in the past. In terms of defense, I would evaluate strength, height, and change of directions, while also considering FG% of opposing team when PF is the primary defender. From the center, I would consider opposing team FG% within 5 feet of the basket, as the center is the last line of defense and percentage of passes successfully secured, as the center is the most likely to be receiving passes underneath the rim for easy layups and dunks. All values mentioned will be considered in comparison to the league average. Because my team is new, the focus would be on winning as many games as possible to develop an enthusiastic fanbase and implicitly indicate to prospective NBA players that my team is seeking to create a winning environment.

See you soon,

Vedant Sachdeva

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.