clock menu more-arrow no yes

Filed under:

Correlating College Stats to NBA Success: Wing Edition

This article takes a look at which stats are the strongest indicators of success for wing players in college.

NBA: Indiana Pacers at San Antonio Spurs Soobum Im-USA TODAY Sports

This article is an updated version of what I did last year—looking at basic college statistics from wing prospects in recent years, and seeing which of these stats correlate most with NBA success (viewed through the measure of advanced statistics). This analysis is not a predictive model. I’m not trying to argue that one player is superior to another just because he has better stats in one area, or worse in another. Prospect scouting is far more complex than that. One of those complexities is the abundance of data on every player coming out of college. I wanted to find a way to filter through some of the noise—discover which stats indicate NBA success. While it is important to consider the whole picture when evaluating each player, I thought that some stats had to be far more significant than others. This hunch certainly seems to be valid.

There are questions about this method of analysis, of course. Most importantly, how does one measure basic stats in college? I use the players’ actual numbers (not per 40 minutes or with pace factored into the equation), as I think that playing time should matter; those who get 30 minutes per game instead of 20 probably deserve them more. For non-freshman, I averaged the stats of their last two years in college. This formula tends to penalize upperclassmen who might dominate their junior or senior years statistically without crippling those who might not have played extensive minutes in the previous years. Finally, are NBA advanced statistics really that viable in determining prowess? To some extent, yes: the five highest Win Shares/48 among wings in the 2010-2014 drafts are Kawhi Leonard, Jimmy Butler, Paul George, Gordon Hayward, and Otto Porter (with Klay Thompson and Jae Crowder close behind). While not the be-all and end-all, advanced stats can provide a relatively accurate approximation of NBA ability: I think that there is value to be found in them.

2010 Wing Draft.csv

Players Age PPG APG RPG SPG BPG TOPG 3PT TS WS/48 BPM VORP/Season
Players Age PPG APG RPG SPG BPG TOPG 3PT TS WS/48 BPM VORP/Season
Evan Turner 21.6 18.9 5 8.2 1.7 0.9 4.4 0.5 0.582 0.051 -2 0
Wes Johnson 22.9 14.5 1.8 6.3 1.3 1.1 2.2 1.6 0.558 0.038 -1.2 0.35
Al-Farouq Aminu 19.7 14.4 1.4 9.5 1.2 1.3 2.9 0.4 0.544 0.067 -0.3 0.74
Gordon Hayward 20.2 14.3 1.8 7.4 1.3 0.8 2 1.8 0.628 0.128 2 2.3
Paul George 20.1 15.6 2.5 6.7 2 0.9 2.8 2 0.579 0.145 3.7 3
Xavier Henry 19.2 13.4 1.5 4.4 1.5 0.5 1.9 1.9 0.591 0.013 -5.1 -0.3
Luke Babbitt 20.9 19.4 1.8 8.2 0.8 0.8 2.2 1.1 0.589 0.065 -2.7 -0.1
James Anderson 21.2 20.3 1.9 5.8 1.3 0.6 2.2 2.3 0.609 0.037 -2.8 -0.1
Elliot Williams 20.9 11 2.2 3.2 1 0.1 1.9 1.1 0.542 0.024 -5.3 -0.2
Damion James 22.6 16.7 1.2 9.7 1.3 1 2.2 1 0.555 -0.001 -4.3 0
Dominique Jones 21.6 19.8 3.7 5.9 1.5 0.5 2.7 1.7 0.54 0.017 -3.6 0
Quincy Pondexter 22.2 15.7 1.7 6.7 1 0.5 1.9 0.3 0.582 0.087 -1 0.3
Jordan Crawford 21.6 15.1 2.6 4.1 1.1 0.2 2.2 1.7 0.549 0.043 -2.1 0
Lazar Hayward 23.5 17.2 1.3 8 1.5 0.4 2 1.7 0.554 0.02 -4.6 0
Landry Fields 21.9 17.3 2.5 7.7 1.4 0.7 2.3 0.9 0.564 0.084 -0.1 0.5
Lance Stephenson 19.7 12.3 2.5 5.4 0.9 0.2 2.4 0.5 0.493 0.074 -0.4 0.5
Devin Ebanks 20.6 11.3 2.5 8 1 0.7 2.1 0.1 0.527 0.035 -4.7 0

This is a sample from the data, the 2010 collegiate wing draft class. Immediately, one can see that Hayward and George are by far the best players by all advanced statistics. They are followed by legit NBA rotation players such as Al-Farouq Aminu, Evan Turner, and Quincy Pondexter. As mentioned above, the advanced stats across the board mostly matched up well with “consensus” and “eye-test”, which was reassuring before the real analysis started. Every drafted wing player who got at least a few hundred NBA minutes was included, even those with extremely short careers such as Dominique Jones and Damion James. Future research might eliminate these players in favor of those with more of a substantial sample size. Then again, their very lack of playing time is also a telling factor, and shouldn’t be ignored.

Last year, when looking at the 2012-2014 data, the three college stats that correlated most strongly with NBA advanced statistics were rebounds, assists, and steals. I was curious if this data would hold up with updated information and two extra classes (2010 and 2011) added. For the most part, the previous research carried held: rebounds and assists were two of the most significant predictors in all three advanced stats, and steals in two of them. However, two more statistics notably joined them-- age and turnovers. True shooting and points per game remained relatively valueless, however, and neither made three-pointers nor blocks per game made much of an impact. This revelation is surprising, as wings are the position most associated with scoring and shooting, and one would think that strong points per game and efficiency in college would translate to future success. Instead, it is the other stats that prove more important.

Theoretically, the model created via regression has *some* predictive value, but I don’t like using it in that way. The primary reason is that prospect placement matters so much for development and success—players might work out beautifully on one team and completely fall out of the league on another. Therefore, it’s very difficult to predict success. Instead, I think of the stats as indicators of potential success. A wing who grabs a lot of rebounds and has a low turnover rate is someone to watch, as is one who has a high assist and steal rate, even if they don’t shoot the ball well. To clarify the model, age and turnovers have negative coefficients (the higher those variables, the lower the predicted advanced stats), while rebounds, assists, and steals have positive coefficients (the higher the better). These results make sense, as youth is generally valued highly in the draft, turnovers are bad, and the other stats are all helpful.

Just for fun, here are the highest and lowest collegiate players in each significant category for the 2010-2014 drafts. Most of the players who were in the top five in any given category have panned out, with some notable exceptions. The bottom fives are rife with duds, though again, not entirely so. The most interesting stat is probably turnovers, with all the players heavy in turnovers (outside of one) becoming good NBA players.

Top 5 rebounds: Andre Roberson, Kawhi Leonard, Damion James, Al-Farouq Aminu, Kyle Anderson

Bottom 5 rebounds: Tony Snell, Joe Harris, John Jenkins, Nik Stauskas, Elliot Williams

Top 5 assists: Evan Turner, Anderson, Iman Shumpert, Dominique Jones, CJ McCollum

Bottom 5 assists: Cleanthony Early, Shabazz Muhammad, Derrick Williams, TJ Warren, PJ Hairston

Top 5 steals: Jae Crowder, Shumpert, Jordan Adams, Chris Singleton, Paul George

Bottom 5 steals: Doug McDermott, Tony Snell, Tim Hardaway, Nik Stauskas, Tobias Harris

Top 5 turnovers per game (lowest): Hairston, DeAndre Liggins, Glenn Robinson III, Reggie Bullock, Crowder

Bottom 5 turnovers per game (highest): Turner, Klay Thompson, Archie Goodwin, Aminu, George

What does all this mean? Well, I think the model eventually makes sense when you consider player development and how wings play in the modern NBA. While “3 and D” players are all the rage in the NBA these days, such players are for the most part very limited. The best of them don’t just hit threes and play defense—they can also move the ball and bring down rebounds and do the little things (think Andre Iguodala or Khris Middleton). Pulling down rebounds isn’t just a function of being large and athletic (though that plays a big role), it’s also about basketball IQ and hustle. Similarly, assists can be manufactured in certain contexts, but generally reveal high IQ and vision. Instead of thinking of steals as a result of gambling (or even luck), they can be seen as usable athleticism and being able to read opposing players well. All these traits can carry over from one level to another, whereas scoring can be dependent on various skills that work in college that might not translate to the NBA. As always, these results are not gospel by any means. They’re just something to keep an eye on as the draft approaches.