With the 2023 NBA Draft coming up later tonight, I thought it would be interesting to take a quick look at some of the players entering the association. As most of the players drafted in the first round tonight will be 21 or younger1The top 16 players in The Athletic’s Draft Rankings are 20 or younger, while 27 of the top 30 are <21 years old. and are far from finished products, player development and fit will have a huge effect on the outcome of each player’s career. Because of this2The small roster size of the NBA and lack of a large minor league infrastructure mean that drafted players generally have to contribute to a certain degree early in their careers. , I’ve never been a big fan of NBA Draft models that generally use a combination of college statistics, athletic measurements and other factors to predict a player’s success in the league. Draft models are definitely useful, but less so in the NBA than in baseball or football3There is also the issue of putting statistics in context and accounting for competition level, which is particularly difficult in a dynamic sport like basketball, where there are fewer individual matchups (like pass rusher vs blocker or pitcher vs hitter)..
Draft models aren’t the only use for a player’s measurables and performance statistics though. Another effective application of this data is the creation of comparable players, or “comps.” Knowing which historical players a potential draftee is similar to can be very useful in determining their fit on a given team, as well as the draftee’s potential developmental roadmap. Along these lines, I built a simple model that utilizes performance statistics4As well as player height to keep the comps somewhat similar in body type to determine 3 comps for the top college players in this year’s draft class. The model uses 15 features in total. These include measures of a player’s shooting (FT%, 2P%, 3P% and True Shooting%), shooting tendencies (FT rate and 3PT rate), rebounding (ORB% and DRB%), ball-handling (AST% and TOV%) and defense (BLK%, STL% and PF/100 possessions), along with a player’s height. Simple Euclidean distance is used to compare each 2023 draftee for all 15 features5The Z-score of each feature is used with a database of 424 college players, all of whom were selected from 2011-2020.
Keep in mind that these are not designed to be traditional body-type comparisons or comps to a given player’s current NBA role. The comps are measuring the similarity of a 2023 draftee’s role and statistics in college with past draftees’ college roles. They are also not attempting to predict a player’s NBA future; comps with good NBA players are nice but not very indicative of a prospect’s future6The opposite is also true of course.. Without further ado, here are my model’s player comps7Consensus rankings courtesy of jacklich10.com/bigboard/nba/:
Overall, I’m pretty pleased with the model’s results. Without being given positional data, or body type data other than height, it did a good job of deducing a player’s position and style. Focusing on the top of the draft, Cam Whitmore is known as an explosive athlete with some questions about his shot, and is quite similar in this way to his top two comps, Kelly Oubre Jr and Justise Winslow. Anthony Black is a tough and athletic guard with shooting questions coming out of college, like his most similar player, Elfrid Payton. The model also did nicely with Gradey Dick, a great shooter and off-the-ball player, comparing him to Malik Beasley and Luke Kennard.
This approach certainly isn’t perfect though. For instance, Brandon Miller, the projected #2 or #3 overall pick, is compared to three second round picks. While these players played in a similar style as Miller in college, they didn’t have the athleticism that Miller possesses8Though it may be somewhat fair to question Miller’s ceiling. Another player that epitomizes the model’s approach is Dereck Lively II. While Onuaku and Christmas are significantly shorter than Lively, they had a similar style of play, with a very high BLK% and few points outside of the paint.
The predominance of NBA backups and role players in these player comparisons is also a sobering reminder of the low success rate of draft picks. For instance, while Elfrid Payton may seem like an unflattering comparison now, he was selected at #10 in 2014 and played in the NBA for 8 seasons. While that outcome would be disappointing for the team that selects Anthony Black, it is far more likely than stardom. In the NBA draft, every successful pick is like finding a needle in the haystack.
One more thing: To visualize the play styles of all 29 college players analyzed, here are some radar charts of their rankings in all 15 model features9A lower TOV% and fewer PF is shown as a data point further from the center.. Swipe through the slideshow to see all 30 players and their profiles:
The code and datasets used in this article can be found on Github.