chapter three

3 A statistical justification for tanking

 

This chapter covers

  • Visualizing means and medians
  • Sankey Diagrams
  • Expected Value Analysis
  • Hierarchical Clustering and Dendrograms

Tanking is the practice of purposely losing games in one season in order to secure a higher pick in the draft prior to the next season. The NBA draft, held annually during the offseason, is where teams take turns selecting the best amateur players from the United States and abroad. Like drafts in other US professional sports, the NBA draft is the exact opposite of a meritocracy; it is a worst-selects-first system that “awards” teams with the lowest number of wins to pick highest in the draft and obtain rights to the best available players.

Our purpose here is to demonstrate that tanking is a rational and worthwhile strategy for teams that are less than successful yet ambitious. Teams tank because it works. Superstars are an absolute requirement to win championships. And superstars are almost always top-five selections in the amateur draft (though top-five selections are not guaranteed superstars). Therefore, teams must lose way more games than they win, intentionally if that’s what it takes, to draft a potential superstar.

3.1 More on tanking and the draft

3.2 Loading packages

3.3 Importing and viewing data

3.4 Creating another derived variable

3.5 Visualizing means and medians

3.5.1 Regular season games played

3.5.2 Minutes played per game

3.5.3 Career win shares

3.5.4 Win shares every 48 minutes

3.6 Preliminary conclusions

3.7 Sankey Diagram

3.8 Expected value analysis

3.9 Hierarchical clustering

3.10 Summary