Skip to content

saurabhr

  • Home
  • Analytics Apps
    • NBA Headshot Plot
    • NBA Player Combos
    • NBA Playoff Matchups
    • NBA Rotation Finder
    • Exploring the EFL Pyramid

About the Project

I was one of 200 stu­dents out of 1000+ appli­cants to par­tic­i­pate in the NBA’s first annu­al hackathon. Over an entire day of cod­ing, myself and two team­mates devel­oped a new met­ric to quan­ti­fy how well NBA play­ers are able to defend shots. Our met­ric was well received — par­tic­u­lar­ly to do the cohe­sive pre­sen­ta­tion of the infor­ma­tion (scope of the work, use­ful­ness, and use of data visu­al­iza­tion) and lead to a top 5 fin­ish in the hackathon

Per­son­al Impact

In my team of 3 I had the most tech­ni­cal expe­ri­ence, con­se­quent­ly han­dled the major­i­ty of the analy­sis (via R) and the data visu­al­iza­tion (R/ggplot2) while my team­mates focused on the pre­sen­ta­tion and quick data analy­sis (via Excel).

Tools Used

Project Detail

Back­ground

Our team devel­oped a new defen­sive met­ric — con­test qual­i­ty, to gauge how well NBA play­ers were at con­test­ing shots, inde­pen­dent of the defen­sive sequence before that.

Intro

We chose to tack­le the fol­low­ing prompt:

Devel­op a new method or tool for eval­u­a­tion of defen­sive per­for­mance in the NBA.

Cur­rent­ly, there are robust ways to detect shot qual­i­ty, such as the Quan­ti­fied Shot Met­ric (qSI)devel­oped by ESPN. The idea behind qSI is that val­ue of a giv­en shot can be deter­mined by shot dis­tance and the dis­tance of defender(s) to the shoot­er. The met­ric we devel­oped, Con­test Qual­i­ty, looks at how well play­ers are con­test­ing shots inde­pen­dent of shot quality.

Process

A shot 15 ft away from the bas­ket with a defend­er 2 ft away from has an expect­ed Pts/FGA of 0.74 (eFG% = 0.37). The eFG% under dif­fer­ent cir­cum­stances is defined by the heatmap below:

Shoot­ers under those aforemt­nioned con­di­tions, defend­ed by Kevin Durant have an Exp. Pts/FGA of 0.68, cre­at­ing an expect­ed points dif­fer­ence of ‑0.06. By tak­ing that expect­ed points dif­fer­ence for every shot a play­er defend­ed over the past two sea­sons (14/15 & 15/16), aver­ag­ing that amount of shots they defend­ed. Shots Durant defend­ed the past 2 sea­sons had an expect­ed 1.000 Pts/FGA, but in real­i­ty Durant only allowed  0.839 points.

Analy­sis

To bet­ter visu­al­ize how play­ers per­formed their charts have been strat­i­fied by role (guard, wing, & big).

A play­er’s pres­ence in quad­rants II & III indi­cates they’re forc­ing low per­cent­age looks.  Their pres­ence in quad­rants I & IV indi­cates they’re con­test­ing shots well.

wing
guard
big

Guards: Redick stands out as a play­er that both forces poor shots and con­tests those shots well. Con­trast that with Con­ley, who does an excel­lent job in forc­ing low­er per­cent­age looks, but does not do a great job of con­test­ing them. This does­n’t nec­es­sar­i­ly mean that Con­ley is a poor defend­er, but that he might be doing a poor job of con­test­ing shots, despite doing every­thing else well.

For­wards: Durant is an enor­mous out­lier. While he isn’t doing the best job of forc­ing bad shots, his abil­i­ty to con­test turns would-be good shots into bad ones. Jabari Park­er is the worst of both worlds — allows good looks, and does a poor job of con­test­ing them after the fact.

Bigs: The big men pass the eye test as well. Dray­mond ranks as one of the best at con­test­ing shots, while some­one like Ran­dolph scores poorly.

Sec­ondary Analysis

Our ini­tial instinct was a cor­re­la­tion between anthro­po­mor­phic wingspan, height, stand­ing reach) and con­test qual­i­ty, and ini­tial analy­sis. How­ev­er, break­ing out the data into play­er roles (guard, wing, big), the cor­re­la­tion is more dubious.

Out­come

Our work received a top 5 final­ist award and was par­tic­u­lar­ly laud­ed for our visu­al sto­ry­telling approach. 

  • Email
  • LinkedIn
  • Github
  • Twitter
  • Instagram
Proudly powered by WordPress
×