Let me take a second to first of all say that its tremendous how the response to our insight articles.
Jose is an amazing sports analytics mind that hails from Madrid, Spain. Hope you enjoy this read.
Once upon a time most of NBA players entered the league after four years developing their game in the NCAA, players who skipped college ball where trivia questions and life was probably easier for coaches and front offices.
Heraclitus, the Greek philosopher, wasn’t good at basketball but he knew his business, and when he said “change is the only constant in life” he was spot on. High schoolers, one-and-done, overseas prospects… NBA teams suddenly needed to develop their young players earlier, when they were full of potential… and question marks. Adjustments take time, and by 2005 the NBA G League (known as NBDL) was born.
In this article, first we’ll check how NBA teams has made use of this new tool, then we’ll evaluate the improvement of players assigned, trying to measure the impact of the G League as a whole, and finally, we’ll try to identify patterns of success; set of players with similary “profiles of improvement” who has had success earning a spot in the NBA.
EVOLUTION OF ASSIGNMENTS
Since 2005, the way NBA teams make use of their affiliate teams has changed, especially from the 2012-13 season when teams assigned players almost 3 times more than in the previous season (note the dot line in the first picture).
On the other hand, if we focus on the 2012-2018 period, we can see that while the number of assignments keeps growing the way NBA teams use the assignments (expressed in the percentage of multiple assignments for the same player) remains stable.
Taking into account the former information, this analysis is centered in the 2012 – 2018 period.
In this analysis. we consider that the improvement due to assignment can be evaluated in the following season, further improvement, two or more years after the assignment, it’s harder to stablish as consequence of the assignment (the player shared more time with the team, had several summers to improve …). We’ll compare every players’ stats in an assignment year against the same indicators in the following year including general stats, shot stats and defense stats.
Example: if a player is assigned in the 2015-16 season, we’ll compare his performance in the 2016-17 season against his performance in 2015-16, this way, we’ll have a set of indicators of his improvement after the assignment. In other words, we’ll store the difference of performance between the current year and the following one.
The G Leagues is a development league, so we consider the origin of the players an important factor; we’ll make distinction of their origin based on the draft (first rounder, second rounder and undrafted).
In order to evaluate whether the indicators are good or not, we’ll add a context, this context will include the performance of the same set of players when they were not assigned. This way we avoid unfair comparison with players that don’t need time to develop and therefore have better indicators.
Following the former example, we compare the difference of performance of the player between 2015-16 and 2016-17 (we could call it improvement due to assignment), against the difference of performance between 2016-17 and 2017-18.
In summary, we’ll have for every season the difference of performance between that season and the following one, and we’ll identify which of that seasons correspond to an assignment as our goal, and which of that seasons don’t correspond to an assignment as our context.
We’ll collect numbers in three groups; general stats, shot stats and defense stats.
PATTERNS OF SUCCESS
With all the data analyzed in the previous sections, we’ve looked for assignments following a similary pattern of improvement through clustering analysis. Then we’ve looked into these clusters trying to identify clusters (set of players) with success in the NBA following their assignment to the G League. If we find that most of the players in a cluster had success in this regard we’ll consider that a “cluster of success”.
In this document we assume that the players who belong to the same “cluster of success” follow a similary pattern, so coaches and scouts can take a look in depth and try to identify factors (out of scope for this analysis) to explain this success, and trying to predict future success for players currently assigned who have characteristics in common with the members of these clusters.
In no way the former definition means that players outside “clusters of success” didn’t have success following their assignment, it just means than most of the players in “clusters of success” had it. In other words, not every successful player is in the clusters, but most of the players in the clusters are successful.
Scope of the report
This analysis is centered in the 2012 – 2018 period for general stats, and 2013-2018 for shot and defense stats.
For general stats, we’ve made clusters based on the improvement of; PTS, REB, AST, OREB_PCT, DREB_PCT, USG_PCT, TS_PCT and AST_PCT. We’ve defined 10 clusters initially and, with the bigger cluster, we’ve done re-clustering and redefined it in 3 new clusters.
For shots stats, we’ve made clusters based on the improvement of; FGA, FG_PCT, EFG_PCT, FG2A, FG2_PCT, FG3A, FG3_PCT. We’ve defined 10 clusters initially and, with the bigger cluster, we’ve done re-clustering and redefined it in 3 new clusters.
For defense stats, we’ve made clusters based on the improvement of; D_FGA, D_FG_PCT, NORMAL_FG_PCT, PCT_PLUSMINUS. We’ve defined 10 clusters initially and, with the two bigger clusters, we’ve done re-clustering and redefined it in 4 new clusters.
The patterns of success identified in this document correspond to the data analyzed in the previous sections.
THE CLUSTERS OF SUCCESS
Conclusion, I feel that Jose, has come up with some great research and some good output… Lets see how many more G-League alumni make it to play top level basketball either in the NBA or move overseas to Europe or elsewhere…