While this post is not about speed climbing, it is about another sport that I follow closely: NBA basketball. About a year ago, I started a side project analyzing 40 years of NBA draft picks and the career performance of those picks while they played in the NBA. It was a fun project that had many twists and turns. At the end, the results of my analysis surprised me.


I liked the project enough to create a summary poster of my work for the American Statistical Association’s Data Visualization Poster Competition. I was excited that I placed in the top five posters in the nation for 2022.


Below is the abstract of my work:


PROJECT OVERVIEW AND QUESTION FORMULATION
While reading The Undoing Project by Michael Lewis, I was fascinated by the storyline of how National Basketball Association (NBA) teams started using advanced statistics around 2000 to better predict the performance of draft picks and potential trades. The teams wanted to take Major League Baseball’s “Moneyball” techniques and apply them to basketball. It made me wonder if NBA teams had improved at selecting draft picks in the past two decades. In a perfect NBA draft, the best player would get picked first, the second-best player would get picked second, and so on. However, as we all know, this is not the reality. Every year, a few players far outperform their lower pick in the draft, while others underperform their high draft pick. But a few outliers do not tell the overall story of how accurate NBA teams are at selecting draft picks. Has draft pick selection accuracy improved during the age of advanced sports analytics?


METRIC SELECTION
The method I used to determine the accuracy of a draft year is by comparing the draft order of players to their career performance rank as measured by Box Plus/Minus (BPM). BPM is a “basketball box score based metric that estimates a basketball player’s contribution to the team when that player is on the court” (Basketball-Reference.com). I chose BPM as the advanced statistic to represent player performance because of a study by DunksAndThrees.com, a professional NBA analysis website, which compared NBA player performance metrics. Based on that study and the fact that BPM data is available for the entire period of the modern era of NBA basketball which started in 1979, it was the best all-in-one advanced performance metric for my analysis.


DATA FILTRATION
Only players who played at least 82 games, which is a full NBA season, in their professional playing careers were included in the analysis. This is because there are a few players who had amazing performances while only playing in a limited number of games, placing them above all-time great players.


PERIOD OF ANALYSIS
1979 was selected as the first year to begin the analysis as this is the beginning of the “modern era” of NBA basketball. Experts consider the early 1980s as the beginning of the “modern era” because the three-point line was added to the game in 1979 and the postseason playoffs were expanded to 16 teams in 1984. I wrangled all data required to do this analysis from the drafts of 1979 to 2019. The 2020 and 2021 drafts were not included in the analysis. 2020 was not included because its total number of players who had played 82 or more games was less than half the median number of players with 82 or more career games across all of the other draft years in the modern era. 2021 was not included because no player drafted that year had played 82 career games.


PEARSON CORRELATION COEFFICIENT (by year and decade)
Linear regression was used to account for a discrepancy in the number of players per year due to the data filtration process. By using the Pearson correlation coefficient, it was possible to compare different draft years that have different numbers of players.


ACCURACY MEASUREMENT
The closer the draft pick order aligns to the ranking of the players’ career BPMs, the more accurate the draft year. The Correlation Coefficient of the BPM rank versus the draft pick order measures each year’s accuracy.


Z-SCORE
The Z-score was used to normalize the data by calculating the distance in terms of standard deviations from the mean. This allows the annual draft year data to be easily compared and shows another graphical view of the data.


DATA SOURCES
Data was collected from Basketball-Reference.com and NBA.com.
Research was incorporated from DunksAndThrees.com.
Player data was collected from the beginning of the 1979-80 season until December 28, 2021.


SAMPLE DATA
Draft_Year 1992
Draft_Pick_Order_Number 1
Draft_Pick_Round 1
Team Orlando
Player_Last_Name O’Neal
Player_First_Name Shaquille
Years_Played 19
Games_Played 1207
BPM 5.1