clock menu more-arrow no yes mobile

Filed under:

Which statistics translate from college to the NBA?

Digging deeper into a sea of amateur and professional statistics

COLLEGE BASKETBALL: NOV 26 Continental Tire Challenge - Duke v Gonzaga Photo by Brian Rothmuller/Icon Sportswire via Getty Images

The NBA is about to enter the final stage of the 2022 season with the Finals just around the corner. For Knicks fans, though, there is an entirely different target date in mind: that of the upcoming draft, which will take place on Thursday, June 23—less than a month from now!

As a loyal reader of Posting & Toasting, you know that we’ve been writing about what happened last year, who did what, which prospects are coming strong, and a wide variety of other offseason topics including the draft lottery and some draft-related research.

About that last point: I’m back at it. Researching and writing about how draft age impacts NBA success—or the lack of it—got me interested in exploring the draft from different angles to try and get some insight into what works, and what does not. At least from a historical perspective, that is, as the draft is a true crapshoot and we all know about that.

I decided to do something very simple: find the correlation between a player’s last college season statistics and the ones he put up in his rookie season in the NBA and playing among pros. Instead of keeping it super simple—single pairs of one-for-one correlations, for example, NCAA points per game to NBA points per game—I built a humongous correlation matrix that links all of the statistics with each other—for example, NCAA points per game to all other 38 metrics present in my dataset.

The idea behind this was to see what clicks at both the amateur and pro levels of play. Are points in the NCAA correlated to rebounds in the NBA? Do prospects with a high amateur TS% go on to put tiny STL% numbers in the NBA? It’s time to explore.

The data set I’m using contains all college/rookie season pairing from drafted rookies in the 2011-2019 span (thus, the rookie seasons took place in the 2012-2020 frame) that went exclusively from the NCAA circuit to the Association.

Basic-statistic correlations between NCAA and NBA play

I guess it makes sense to start at the beginning. That is, when it comes to basketball, looking at the simplest of statistics that have been around forever: minutes, points, rebounds, assists, steals, and blocks per game.

Just as a quick primer, the correlation between two variables ranges from -1.0 to +1.0. A perfect negative correlation is represented by the value -1.0, while a 0 indicates no correlation, and +1.0 indicates a perfect positive correlation. A perfect positive correlation exists when one variable decreases as the other variable decreases or one variable increases while the other increases. A perfect negative correlation means the relationship that exists between two variables is exactly opposite all of the time.

With that in mind, here’s the correlation matrix for those six classic metrics.

Here are some takeaways from that chart:

  • Blocks have the strongest one-for-one correlation between NCAA and NBA outcomes, at a high 65%.
  • Assists are a close second on that front, with a high 61% correlation between the last NCAA season and the first NBA campaign from rookies.
  • No other metric is above 50%, with steals ranking third but already “down” at 47%. None of the three metrics show incredibly strong correlations (below -70% or above 70%) but all of them are moderately high (between 30% and 70%).
  • Most interestingly, though, are the correlations between seemingly unrelated metrics. The more NCAA minutes per game, the fewer NBA blocks and rebounds per game. Great NCAA rebounders turn out to be great NBA rebounders... and fantastic pro-shot blockers, too. Something similar can be seen on the assists/steals front, as the correlation is approaching a positive 40% (37%).

Shooting Correlations

Instead of looking at per-game statistics, which operate on very different stages in the NCAA and the pro game and are massively impacted by usage and roles, it makes sense to look at percentage and rate metrics for a better picture of what can be expected once prospects make the jump to the highest of levels in joining the NBA.

I am starting with some shooting metrics, including the classic ones (FG%, 3P%, FT%) and then some other fresher calculated stats (TS%, eFG%) that offer a better image of a player’s shooting prowess. Here are the results.

And again, some quick takeaways:

  • Good shooters tend to be good shooters no matter the level they play at. Good FG% stays up jumping from the NCAA to the NBA, and that 40% positive correlation is the strongest among every shooting metric depicted in the chart above.
  • Something similar happens with three-point shooting (36%), although there is quite a bump down in correlation from free-throw shooting in the amateur ranks to the marks posted in the NBA at just 27% (already into weak-correlation territory).
  • Interestingly — and this has been widely discussed around scouting corners — amateur FT% tends to correlate nicely with professional 3P%, the relation sitting at 27%. That’s nothing otherworldly, but it is the largest of correlations between seemingly unrelated metrics among those used for this category.
  • When it comes to negative correlations, FG% and 3P% are the two most-clashing stats in terms of NCAA and NBA production. The higher field-goal percentages in the NCAA, the lower 3P% in the NBA and vice-versa. Something similar happens with FT%—the lower scoring from the charity stripe in the NCAA, the higher FG% in the NBA, and vice-versa.

Rate Correlations

I also thought it’d be interesting to look at some rate statistics to avoid usage/minutes bias and data skewing. That means that instead of looking at just, say, assists per game, I’ll be exploring statistics such as Assists Percentage or Rebounding Percentage. Those metrics are calculated as the estimate of the percentage of teammate assists/rebounds/etc... that a player got while he was on the floor.

For this, I’m using the classic rebounds, assists, steals, and blocks, along with turnovers and usage rate. This is the resulting correlation matrix.

Some takeaways from the best-correlated set of statistics:

  • In comparison to the basic stats and the shooting metrics, this little data set of percentage stats is the one showing the strongest correlations between amateur and pro play. It makes sense, considering that (for example) the best rebounders of an NCAA team are probably still going to be the best rebounders of an NBA team.
  • That is precisely the case with rebounds (73% positive correlation), assists (72%), and blocks (70%) on a one-for-one basis. It was expected, and the correlation can’t really be much stronger than it’s already been in the past 10 years of data.
  • Interestingly, the steal percentage doesn’t show such a strong correlation between the amateur and the pro game, though. Although the correlation is positive (expected), it sits at a very low 23% compared to the other three main metrics all being at 70%+.
  • Steal percentage is the one metric of the bunch that has a better correlation with another metric than with itself when translating marks from the NCAA to the NBA. It is close, but it has a stronger correlation with AST% (28%) than with itself (23%)—the more steals in the NCAA, the more assists in the NBA and vice-versa.
  • The negatively correlated metrics show how the jump from the NCAA to the NBA means players tend to specialize more once they turn pro. That is most notable in the rebound and assist categories: players with too high or too low TRB% and AST% have the largest negative correlations between those two stats going from one stage to the other. That can be understood as NBA teams limiting players who grab a lot of rebounds to focus on that, thus dropping their assist numbers, and also the opposite (great NCAA assisters just get the task of grabbing boards removed from their to-do list once they join an NBA team).
  • Highly used players in the collegiate circuit tend to retain similar roles in terms of usage rate in the NBA. Of course, the correlation is barely moderate (32%) because rookies are just that, rookies, while they were the bona fide studs and stars of their NCAA teams. But this clearly shows a high enough correlation to tell how players highly regarded as amateurs tend to be those most used in their first season as NBA players.

Offensive/Defensive/Overall Contribution Correlations

Finally, I have calculated and plotted the correlation between Offensive and Defensive Rating, and PER in the NCAA and its NBA translation. Here are the results.

And your quick takeaways:

  • The DRtg metric correlates much more strongly than the ORtg one. Offensive contributions have a positive 13% correlation between the amateur and the pro levels of play, while the defensive ones have a much stronger 28% (more than double) correlation between both leagues.
  • The ORtg of prospects in the NCAA correlates positively with its NBA counterpart and doesn’t have anything to do at all with NBA defensive ratings (random/no relation at all, 0%). That’s not the case with DRtg, as it has the strongest one-for-one correlation but (even on a smaller percentage of 8%) correlates negatively with offensive production (that means the higher DRtg in the NCAA, the lower ORtg in the NBA and vice-versa).
  • While not the most trustworthy metric to measure the overall impact of players in the basketball game, PER is a good-enough stat to quickly know who’s contributing the most to their teams. Turns out (as should be the case in most cases) that the best prospects balling in the NCAA go on to do so in the NBA too. Too bad, again, knowing the NBA draft is truly a crapshoot so that correlation figure is just at 30%.

Full Correlation Matrix

For all you number nuts out there, here is the full matrix. Definitely click on it to open it in full size, or just save it to your device to better explore it—if you can, which I’m not convinced is even possible.

* AVG represents the average correlation of the NCAA metric with all of the NBA metrics. ABS represents the absolute value of the AVG figures.