Explores whether "Hustle Stats" have any correlation with wins/player value
To view the full write up: https://github.com/tbala25/R_NBA-Grit/blob/master/BALA_T_report.pdf
This report will explore professional basketball statistics to try to find how much truth is in the day old mantra ringing through gyms across the country, “hustle wins games.” I have recorded a journal throughout this process so you will be able to read about my trials and triumphs with each step along with the results and justification for that step.
The data was scraped from the National Basketball Association’s statistics portion of the site. I used import.io to scrape data from three different tables: Traditional Team Stats, Player Hustle Stats, and Player Tracking Rebounding Stats. Since the question was about hustle on the court the Hustle Stats table was an obvious and was the NBA’s best collection of data for an otherwise immeasurable characteristic. I added the Rebound Table because I felt that the statistic for Contested Rebounds also demonstrated “grit”.
When I first scraped the data I was just playing with coming up with an all-encompassing metric for hustle. I created arbitrary weights for each statistic based on how often I felt it could happen, how impactful it is in the game and trying not to overweight the work of a forward or center versus a guards. I came up with this formula for GRIT:
GRIT = (Screen Assists) + 2(Deflections) + 3(Looseballs Recovered) + 3(Charges Drawn) + (Contested 2FG) + 2(Contested 3FG)
I also decided to make a normalized GRIT statistic which projected each players’ GRIT score as if they played a full 48 minutes. In retrospect this number should have been far closer to 30 which is around the average minutes that starters will play in a game. Originally my GRIT score calculation was incorrect and I could tell because naturally a starter and a big-man would have a higher GRIT score than a reserve guard and this was not true.
