Projecting Draft Prospects Three-Point Percentages


Mandatory Credit: Rick Osentoski-USA TODAY Sports

Last year I made a promise to myself to not to focus much on any NBA draft prospects’ three-point percentage. Not because I long for a return of the ’90’s post up basketball, don’t believe in spacing or bemoan the lost art of the mid-range jumper, but because the percentage a prospect shoots on threes in college or European ball is simply not a very good predictor of how he will shoot in the NBA.

In the models I ran last year, I found that free-throw percentage and three-point attempt frequency were at least equally predictive as the player’s three-point percentage. The models are only modestly successful at predicting, but are still a significant improvement over just looking at three-point percentage.

In keeping that promise, I ran some of this year’s top prospects through the models to ground my expectations in looking at their shooting prowess. Given the variability inherent in shooting from distance, the model ends up producing a fair amount of simply regressing to the mean, however, there is more to it than that. By pulling out tiers of shooters by current three-point percentage the effects of the other model inputs, three-point frequency and free-throw shooting becomes more apparent.

Below are the better three-point shooters among the top forty or so prospects, using Draft Express as a filter, with both of the models I developed for three point percentage and one developed for number of made three-pointers per forty minutes[1. The made three pointers model had a .63 R^2 in sample, factoring free throw percentage, made three pointers pre-NBA per 80 possessions and a small age factor, the details for three point percentage are in my original model last year linked here.]:

In the table, we can see that even though Justise Winslow and Kevon Looney have approximately the same three-point percentage in their college seasons as D’Angelo Russell and Devin Booker, their projected percentages are quite different. Winslow and Looney project as slightly below average three-point shooters while Russell and Booker project as better and more prolific three point shooters[2. If they stick in the NBA, that is, both models were trained necessarily on prospects with enough time playing in the NBA for their numbers to stabilize to a degree].

There are similar results with the middle-tier shooters in terms of pre-NBA three-point percentage. Sam Dekker and R.J. Hunter shot a similar percentage over their college careers but Hunter’s frequency and high free-throw percentage help him project to be an above average three-point shooter over the long haul, while Dekker projects as below average:

As stated above the model itself is noisy, leaving plenty of room for scout and shot doctor evaluations[3. I have been on the look out for contemporary shooting evaluations for past prospects to test in a model]. Some of the players on this list won’t see the floor long enough to evaluate their NBA level shooting prowess to any degree, but here’s the projection for the entire list with a free-throw percentage model thrown in for good measure: