Joseph Leung

Joseph Leung

Research Professional

How much does the NBA Summer League matter to players? (joint with Shivram Vishwanathan)

5 minute read

Published:

Many do not pay attention to the summer league in years past, but as more and more one-and-done NBA prospects enter the league, the coverage of the NBA summer league has gone up. We were curious to see how the NBA Summer League contributes to the development of young NBA prospects in their rookie seasons. While our observed effects may not be causal, we hope to shed some light in how useful the NBA Summer League is for the player and how closely we should follow during the offseason.*

Introduction

The NBA Summer League is the first NBA action we get before the NBA season gets underway. For rookies, the Summer League is either a chance for lottery picks to showcase their skills or a chance for undrafted picks to prove their worth. For older players, it is a chance for them to continue to improve their game and develop in preparation for the NBA season. Due to potential risks for injury, these games are primarily limited to developing players that don’t usually see a lot of playing time in the NBA season, rookies, second-year developing players, G-league affiliate players, and international players trying to break into the NBA.

Because it is hard to estimate the long term effects, we limit our analysis only to players that play in the Summer League and the NBA season directly after the summer league.

Data

There are two main data sets we will use: NBA Summer League and the corresponding NBA Regular Season. For summer league, we will only be using the NBA Las Vegas Summer League data, the largest and highest profile of all the summer leagues. For ease of interpretation, we keep the basic NBA metrics (i.e. ppg, rpg, apg, spg, bpg, fgm, fga, fg%) for both the NBA Summer League and NBA season. Furthermore, we also keep information regarding the player, such as position and minutes played. We start with 3,873 player-year NBA season observations from 2010-2018 and 1,801 NBA Summer League players in 2010, and 2012-2018. 1

We merge the two data sets together on the players that played in both the NBA Summer League and the corresponding NBA season. For example, we merge the 2016 NBA Summer League data with the NBA 2016-2017 season data by player name. Our final sample contains player-level observations that observe a player’s characteristics and summer league and regular season statistics for a given year. This will also give a first cut of the quality of players, since only the best of the NBA Summer League make the NBA team. Due to data availability, our sample will span 7 NBA season - NBA Summer League pairs. Our final panel contains 595 player-year observations.

Analysis - Under revision

To start, I look at the summary stats:

  • Summer League: Points: 11, Rebounds: 4.5, Assists: 1.7

  • Regular Season: Points: 5.3, Rebounds: 2.3, Assists: 1.1

This makes sense: rookies and prospects tend to get less playing time in the NBA season vs the NBA Summer League. Next, I build scatterplots on these metrics to check for association between the NBA season and NBA Summer League. The following are the results:

Points

Points

Rebounds

Rebounds

Assists

Assists

When I run these as regressions, I get significant results. However, aggregately, there is a lot of noise since I am not accounting for a given prospect’s position or the number of games the prospect has played during the regular season. Here’s why this may matter:

  • Centers tend to get more rebounds and Guards tend to get more points or assists. This should be accounted for, but we want to make sure of the variation within each position.

  • Some prospects only play during garbage time, while others play during crucial parts of the game (i.e. starters, role players, etc.).

Here are the next steps for analysis:

  • Control for within position and across position variation. This will help see whether the summer league is associated with a prospects season averages when compared to other prospects that play the same position.

  • Take inactive rookies out of the dataset. Set a minutes per game (mpg) or games played (gp) threshold to ensure that the prospects we look at play similar amounts between summer league and the regular season.

  • Extend the analysis across many years.

  • Try to find causality (extremely hard to do).

Conclusion

In conclusion, I think this is a pretty interesting problem to look at. To sports fans, the summer league is pretty insignificant part of the season, but looking at these trends may help us explain why certain prospects have really good regular seasons and why others have subpar or disappointing regular seasons.

* First drafted: 06/01/2018

1 NBA Seasons data comes from Basketball Reference. NBA Summer League data comes from RealGM. The NBA did not do a Summer League in the 2011-2012 season due to the lockout.

You May Also Enjoy

This NY/China flat is worth HOW much?

8 minute read

Published:

I had a cousin ask me about why Asian housing prices are so much higher than western housing prices (like in New York). Currently in my life, owning real estate does not seem so feasible, so naturally, I have no idea what these prices looked like other than being way above my pay grade. That being said, I investigated the time series prices of housing (more specifically growth of property prices) in New York vs those in China. Furthermore, the higher prices of property observed in China, also extends to Hong Kong.

FRBNY RA: Do all former RAs get PhDs?

14 minute read

Published:

It has been two months into my new job at the Federal Reserve Bank of New York (FRBNY). As a research analyst (RA), I primarily assist economists with creating paper figures and tables, running regressions, and cleaning data.1 It is common knowledge that many RAs move on to pursue a PhD since, in most cases, the RA job is meant to prepare for the things one is expected to do in a PhD program. I pose the following questions:

Markov Chain Monte Carlo and the AUDL Project: Reflections

7 minute read

Published:

For my mathematics senior capstone project, I examine the robustness of a Poisson Goal-scoring model proposed by Everson and Goldsmith-Pinkham (2008). My expository capstone paper is found here. The purpose of this paper is to highlight the goal-scoring model, simplify it, and apply it to a similar scenario. In this blog post, I go back and examine the failure of prediction in my simplified Markov Chain Monte Carlo (MCMC) algorithm. I will go look at two division in particular: the Midwest and West divisions.

Swarthmore College: Graduation

6 minute read

Published:

Today I graduated from Swarthmore College! While this isn’t an academic post, I do have some thoughts to share regarding my time at Swarthmore College. If there are any edits/additions you want to see, let me know.