Analytics Testing: Looking at Prospects by RAPM
Do deep dives into a heavily-filtered plus-minus metric provide accurate views of the type of pro a prospect can become?
Let’s talk analytics. The eye test is something we spend a lot of time with there, diving into the film and using our coaching experience to point us into directions on prospects’ upside and impact. We need the numbers to balance us out and help point us in different directions. As Fran Fraschilla says, the stats indict, the film convicts.
One of the best individual metrics out there for cross-comparison can be Regularized Adjusted Plus/Minus (RAPM), a variation of the standard Plus/Minus stat. The net score while a given player is on the court is defined as his Plus/Minus. If the game is tied 75-75 when Player X enters, and he leaves with the score 90-85 in his team’s favor, his plus-minus for that period of time is a +5.
Plus-minus stats don’t account for competition or other players on the team. You can put nearly any player as the fifth in a lineup of Steph, LeBron, Kevin Durant, and Giannis, and that player will have a positive Plus/Minus. Similarly, your backups matter. The presence of star players can also create biases toward role players who spend a lot of time on the court with them, viewing a particular player's performance as disproportionately productive due to the large-scale impact of the star. Playing on a team with a really poor backup, or where there’s a major glaring hole with the other players in that backup’s lineup, the individual Plus/Minus stats will change.
Individual stats and metrics in a five-player game can be flawed. That’s why they need to be adjusted, or filtered, for the purpose of getting a more accurate reflection of the player while accounting for the variables around them. The best way to do that: look for an average of all the players at a position and compare the player’s impact to that mean, not to their teammates and lineups. Adjusted Plus/Minus ratings do not reward players simply for being fortunate to be playing with teammates better than their opponents.
While that’s an explanation for Adjusted Plus/Minus, the Regularized Adjusted Plus/Minus (RAPM) adds another filter on top of that adjustment to weed out outliers or errors that can come from smaller swings in minutes.
Let’s dive into some of the RAPM numbers for the college players in this draft class. What might the numbers suggest? Where are they off? What statistical filters do they indicate for prospects? And, perhaps most enjoyably, how do we compare it to some historical factors. We have a list of 36,000 data seasons on RAPM since 2013, and can really compare serve as a baseline for comparison.
Top Overall RAPM
Different publications will calculate RAPM differently. We took offensive and defensive metrics, separated them, then added them together instead of calculating everything out as one number. The difference is negligible, but there is a standard error of about 0.15.
The list of the top-eleven prospects here is pretty damn diverse. There are some who are carried by one end of the floor, but in order to get up to this level, metrics have to be at least passable on both ends of the court.
Historically speaking, guys who get north of 10 on RAPM metrics pre-draft are first-round selections. In thinking about the players whose metrics benefit from being on successful teams, 10 of the 11 here made the NCAA Tournament (Vince Williams from VCU is the lone exception). Two were national champions (Ochai Agbaji and Christian Braun). More interesting to us: only four players — Chet Holmgren, Paolo Banchero, Kennedy Chandler and TyTy Washington — were freshmen this year.
Before we extrapolate too much from these numbers, let’s dive into the deeper and more nuanced metrics out of RAPM. For now, this is a solid baseline group of analytic high-achievers.
Top Offensive RAPM
What immediately jumps out are the monstrous numbers from Keegan Murray and Dalen Terry. Murray and Ivey are two offensive hubs coming in at the top of this draft, indicative of their efficiency at such high volume. Terry is a more unique top guy here, but analytic models love him for the high playmaking frequency, how he thrives in transition and sturdy spot-up shot-making.
The numbers at the top of this class are truly elite by this metric. Here’s a list of the five best prospects to be drafted since 2019 by offensive RAPM during the year they were drafted:
Devonte Graham, 2018 (8.26)
Davion Mitchell, 2021 (8.02)
DeAndre Hunter, 2019 (7.77)
Jalen Brunson, 2018 (7.73)
Mikal Bridges, 2018 (7.68)
Murray, Terry, Ivey and Agbaji are all bordering on elite offensive RAPM numbers, with Terry in particular standing out as a high-value role player. None of the guys mentioned above are really good stars, but turn into efficient role players in what they’re asked to do. The best and most efficient number from a 20 PPG scorer in the NBA over that time: Trae Young (7.34).
A few other interesting notes:
Teammates Chet Holmgren & Julian Strawther have the same number. Much of that has to come from them sharing the floor. It’s the perfect case for the star carrying the role guy we have in this year’s class. Take a guess who is the star and who is the role player riding the coattails.
EJ Liddell and Kofi Cockburn, two great post-up players in college, are pretty high on the list. Perhaps the main appeal of their offense is less appealing when placed in an NBA context.
Agbaji and Griffin, two of the best catch-and-shoot guys in the draft class, are high on this list. We get the sense that C&S numbers are looked upon nicely in these analytical models because they sustain no matter who the teammates on the floor with them are.
It’s funny to see AJ Griffin on this list and not Paolo Banchero. In fact, Banchero is third of all Duke prospects here, as Wendell Moore had a higher offensive RAPM than the potential top pick.
Lowest Offensive RAPM
We won’t spend that much time on the lowest guys on the list, but moreso use this as a means for describing different types of prospects. We’ll spend more time on the context than the numbers themselves.
There are those who are really good creators in poor systems or with poor teammates. Bryce McGowens is one of them. The Nebraska roster struggled around him, and while he put up good raw numbers, he wasn’t efficient from 3 and his teams provided very little spacing around him.
There are those who are casualties of the rosters they played on, like perhaps Josh Minott or Peyton Watson are. Neither can shoot, which drags down their individual efficiency, and caused their teams to move away from them during stretches of the season. UCLA already had star players who needed spacing around them (mainly Jaime Jaquez, Jules Bernard and Johnny Juzang), and Watson didn’t fit, so his plus-minus numbers reflect the spacing struggles of the other three when he was on the floor. The same goes for Minott, whose best teammate was Jalen Duren, a big man who scored out of post-ups and struggled have adequate spacing for them when Minott was on the floor. Michigan’s Moussa Diabate had similar issues next to Hunter Dickinson.
All players except one on this list were freshmen. Where the numbers can fail is through equating efficiency with upside. These players have to be good enough to earn a major role, but if they are not surrounded by strong talent (McGowens, Mohammed) or trying to fit in on a contender (Watson), their efficiency can really drop.
That doesn’t mean none of these guys are able to have a positive offensive impact in the NBA. Ziare Williams posted a -1.65 offensive RAPM last year, Josh Christopher a -0.23, and back in 2018 Michael Porter Jr. had a -0.95 RAPM in only 3 games. These numbers speak to production but not to talent.
Top Defensive RAPM
This was the listing that got a great deal of attention. Say hello to Chet Holmgren, a dominant defender in almost every metric and way. With such a monstrous number at 7.18, Holmgren’s defensive impact is legendary. Our RAPM database includes over 36,000 entrants since 2013. The only other Defensive RAPM in the database over the last four years: Xavier Tillman (7.25) and Matisse Thybulle (7.26). Top that 7.0 mark in college and it’s almost a guarantee to be a great defender in the NBA.
Some other comparison numbers for context. Last year Evan Mobley was 3.92 on the scale. Some other recognizable names in the last few drafts who logged above a 5.0 defensive RAPM: Udoka Azubuike (6.98), Franz Wagner (6.82), Jevon Carter (6.24), Brandon Clarke (5.71), and Grant Williams (5.02). Seven players in this class registered a Defensive RAPM over 5.
An interesting name on the list here was Paolo Banchero from Duke. We won’t dive a ton into what that means because, honestly, they’d all be hypotheses and shots in the dark. What it does, however, is serve as a baseline to share that Banchero shouldn’t be considered a negative defender.
Lowest Defensive RAPM
This list tends to line up with the eye test quite a bit. Baldwin, Juzang and Branham were some of the most difficult defenders to watch on film. Terquavion Smith was on a poor team that couldn’t get a ton of stops, and he struggled physically as a freshman. Ivey’s Purdue team was pretty poor on that end of the floor, as was Wesley’s Notre Dame team.
Historically speaking, Baldwin’s number is incredibly poor. He achieved that mark while playing in only 11 of his team’s 32 games — roughly 34% of Milwaukee’s contests. Multiply his Defensive RPM out amongst that full slate of games and he would be on the wrong side of -4.0. No player has ever been drafted below that mark.
The lowest mark of a player during their draft year since 2013: Tyler Harvey in 2015 (-3.35). But it’s incredibly rare to see someone below 0 get drafted in the first round. TJ Leaf went there in 2017 despite a -2.13 Defensive RAPM. Cam Thomas had a -1.43 score last year, and Jerome Robinson had a -0.37 DRAPM in 2018. The defensive litmus test is somewhere around zero, but is not an absolute.
Also, for what it’s worth, Ryan Rollins posted a -2.73 mark his freshman year, which would have been by far the worst in this class. That number shot up to 2.77 this year, another indicator that team success swings these numbers as much, if not more, than individual gains.
Freshman Jaden Ivey had a -2.09 mark, and posted a -0.33 this year. Is that sample size a coincidence, a byproduct of Purdue’s poor defense, or indicative of some real flaws on that end for Ivey?
Well Above-Average on Both Ends
The last time we checked, the game is played on both ends of the floor. Guys who are good on both ends — and far above-average in their class — are the type of players we’re attracted to and are immune to wild swings based on their teams being elite on one end of the floor but not another. Five guys in this class finished with a RAPM above 4.0 on both offense and defense.
Over the last four years, only a few prospects from each class exceed 4.0 on both offense and defense:
2021: Isaiah Jackson (4.07, 4.30), Jason Preston (4.49, 4.08), Herb Jones (4.38, 4.01)
2020: Devin Vassell (5.08, 5.68), Xavier Tillman (6.51, 7.25), Udoka Azubuike (4.07, 6.98)
2019: Zion Williamson (6.98, 4.04), RJ Barrett (6.68, 4.38), Jarrett Culver (4.12, 4.85), Brandon Clarke (5.16, 5.71), Grant Williams (6.09, 5.02), Nic Claxton (4.11, 6.41), Bruno Fernando (4.21, 4.65)
2018: Wendell Carter Jr. (4.51, 4.93), Grayson Allen (7.00, 4.07), Jevon Carter (5.7, 6.24)
Having five this year — pretty damn good, but not necessarily an indication of stardom. How guys like Paolo Banchero and Chet Holmgren fit in with this model definitely remains to be seen.