That assertion is neither a scout’s opinion nor a writer’s hot take, but rather the conclusion of ESPN’s Draft Analytics model, which has been expanded in 2018 to include more information than ever before.
The former Arizona center and the Slovenian point guard are widely regarded as the primary candidates for the Suns’ selection at the No. 1 spot, but the model — based on a variety of criteria — projects Bagley to have the best average early career of anyone in the class. It also believes Bagley has the best chance among this year’s crop of playing at an All-Star level in his first five seasons.
You can read more about ESPN’s model here, but the Cliffs Notes version is that it considers, along with basic information like position and age, data from up to five categories for each player:
We created a model for each of those five categories and then one overarching model that produces our full projection. So what vaults a player like Bagley over his peers? Well for starters, this is not a unanimous decision: Ayton and Doncic both rank higher in ESPN draft analyst Jonathan Givony’s Top 100 (what we use for scout rankings), so Bagley had to make up ground in other areas.
And that he did.
First, according to our numbers, Bagley was a touch more productive in college than Ayton, though the two players were remarkably similar statistically.
Both players were extremely valuable to their teams on the offensive end of the court, but in terms of opponent-adjusted defensive rating, Bagley ranked in the 91st percentile among D-I players last year, while Ayton ranked in the 78th. Keep in mind, this is opponent-adjusted rating, so the fact that Duke played tougher opponents than Arizona is factored into the equation here. Offensively they were similar, though Bagley was superior at drawing fouls.
Ultimately, the difference is minor: Bagley ranks fourth while Ayton ranks seventh in our NCAA production component.
Where the real separation occurs is actually prior to when either was in college: it’s their AAU numbers. Bagley, playing for his father’s Phoenix Phamily team, put up better numbers in most parts of the game relative to Ayton on California Supreme.
Bagley shot better and rebounded and blocked at a higher rate than Ayton. Ultimately, Bagley’s AAU box plus-minus ranked in the 93rd percentile of players on his circuit while Ayton was in the 82nd percentile, though it is worth noting that Bagley was in a much higher usage environment than Ayton at that time. Our AAU component ranks Bagley as the third-best player in this class, whereas Ayton is only 13th.
But that information is old, you might be thinking. What do we need those numbers from before college when we have the data from when they were in college? Because one college season isn’t a particularly large sample, and looking back a little further increases our information pool. This is something not only we do in our model, but a task some NBA teams are undertaking as well.
And how about Doncic? That’s a little more complicated. Though he doesn’t overlap with Bagley in any category other than our scout rankings, we’ve learned enough about the translation of international statistics — and have the benefit of having some players in our dataset who have played both internationally and in college — that we can get a sense of how everyone stacks up. Doncic was exceptional in likely the best non-NBA league in the world, but based on his performance at Real Madrid and what we know about Bagley, the Duke star moved ahead of the Slovenian at the top of our overall rankings. Ayton ranks No. 3.
So where does the model come down on some of the other top names in this year’s class?
Overall, the model likes Jackson in spite of what it would consider a lack of star potential. The system believes he is remarkably solid and the most likely player to end up a regular starter — but not an All-Star — in the NBA with a 42 percent chance. His shot to reach that next level — of being an All-Star — is much lower (8 percent) than someone like Mikal Bridges (15 percent). For teams picking this high, it’s hard to pass upside in a star-driven league like the NBA.
Besides Bagley at No. 1, Bridges is a fairly big call for the model at No. 5. It’s only a handful of spots ahead of where many expect him to go, but this early, that makes a big difference. To wrap up the thought above, Bridges’ 15 percent All-Star chance makes him second-most likely in the class, behind only Bagley, despite the fact that he quite possibly will be the oldest person selected in the lottery. This is all about his college production: At Villanova he was an efficient scorer with strong ball security and provided both steals and blocks en route to a strong overall defensive rating. From the model’s standpoint, looking at his college play, what’s not to like?
Smith, an athletic wing, is ranked worse than ninth in our scout, combine and AAU models but makes up for it with his NCAA production. Offensively he gets credit for scoring efficiently in a very tough conference — though notably with a much lower usage rate than some of his peers — and for contributing on the offensive glass. But our model likes him for his play at both ends of court, and on defense he produced steals and blocks en route to a strong individual defensive rating.
There’s no boom-or-bust prospect in this year’s class quite like Young, the divisive player who took college hoops by storm at the beginning of the season before struggling in the second half. Outside of Lonnie Walker IV, whom the model despises, no one has a higher bust percentage (23 percent) than Young in Givony’s top 15. But at the same time, the former Oklahoma point guard also has the fifth-best shot at becoming an All-Star (12 percent).
Another reason for the inclusion of more information this year: Without it, we would have little to go on for someone like Porter, who missed almost all of the Tigers’ season due to injury. Porter was ranked sixth in this draft class for his youth performance, but only 15th in combine measurements due at least in part to having a below-average wingspan for his height.
Let’s identify some potential values in this year’s draft. Using Givony’s rankings as a proxy for the NBA’s consensus, we can find some players about whom our model strongly disagrees with the general opinion.
Rated 24th by Givony, Melton is the seventh-most likely All-Star in this year’s draft, according to our model. While an efficient college player, his projection based on the combine and AAU rank sixth- and fourth-best in this draft class, respectively. Although he is 20 years old, our numbers suggest he has a similar skill set to Kyle Lowry and a similar projection to what Terry Rozier, who ranked 14th in our 2015 projections, had using this system.
He’s 22, and if you’re going to be an elite NBA player, you don’t typically enter the draft two years into your 20s. Some guys develop later, and Bates-Diop had a phenomenal year at Ohio State in which he ranked highly in scoring efficiency, took great care of the ball despite high usage and was among the better shot blockers in college. His body measurements are easy to overlook compared to the record-breaking wingspan of Mo Bamba, but he has low body fat and good strength for someone of his size, setting him up favorably for facing bigger bodies in the NBA.
Another Greek Freak? While the younger brother of the Milwaukee Bucks superstar did not make a ton of noise in college, he quietly has top-30 projections in our NCAA, combine, international and AAU models (and 31st in our overall model). Kostas is an efficient scorer and elite shot-blocker; perhaps he’ll also end up a late bloomer like his older brother, Giannis.
If our NBA draft model had emotions, it would be utterly perplexed by the consensus feeling on Walker as a borderline lottery candidate. Aside from the scouts’ evaluation of Walker, it sees no empirical evidence that suggests the former Miami guard ought to be selected that high. He ranks 62nd in the NCAA production model, with opponent-adjusted shooting and assist numbers that don’t move the needle. And he had no real impact on the boards. Our AAU model thinks he’s fine (rank: 25th) and his combine measurables don’t jump off the page (20th), so our overall model finds it hard to believe he should be drafted in the teens.
Elie Okobo, PG, France (Analytics: 73, Scout: 21)
In order for our model to think a player from the French LNB Pro A — which Fran Fraschilla ranked as the seventh-best international league back in December — is worthy of being a first-round pick, the prospect in question would likely have had to dominate the French league. Okobo did not. There was some production as a playmaker and he has the ability to shoot from beyond the arc, but he did not blow away his competition the way Doncic did in a better league. Add in the fact that Okobo is nearly 21, and our model remains skeptical of his translation to the NBA.
Though there are aspects of Holiday’s game that are appealing — his ability to score from the field including from 3, where he is an efficient shooter — there are some concerns for Holiday on both ends of the court. Though he may turn out to be an effective defender, that skill did not translate to his numbers last season, where he was middling in steals and had an abysmal individual defensive rating. To be fair to Holiday, the porous defense surrounding him may have played a large role in those numbers. But offensively, he turned the ball over quite a bit, especially considering that the Pac-12 defenses weren’t nearly as tough as those of, say, the Big 12.
Our projections try to estimate the distribution of possible Box Plus-Minus (BPM) values in a prospect’s second through fifth seasons in the NBA. This covers the four years of team control for each rookie and the first year of his second contract. Why omit rookie years? They are often outliers in a player’s performance. For each player we have an average projection, and the chances a player becomes an All-Star, starter, role player, or bust in the NBA. These breakdowns derive from historically looking at how good the BPM in this period was for future All-Stars, starters, role-players, and NBA busts. Based on career lengths for each of these categories, there are about 2.7 All-Stars per draft, 13.5 non All-Star starters, and 24.4 role/bench players.
Last season we had a model that combined scouts’ rankings with a player’s pace and opponent-adjusted statistics from the last two seasons of his college career. For international players we had a similar model that used the scout’s rankings and his box scores from playing overseas. One problem with the international players is that with so few data points in comparison, it is hard to have very good predictions. That is why this year we made five different models and then predictively averaged the results for each player to come up with an aggregate prediction. Some quick details on information used in each model:
Scout rankings are based on ESPN draft expert’s rankings (Chad Ford 2000-’10, Jonathan Givony 2011-’18).
The NCAA and AAU/FIBA Juniors models considered opponent-adjusted per possession box score statistics and composite statistics like individual rating and win shares.
International statistics include the information above and consider the strength of the league.
The combine information is based on body measurements like height, weight, wingspan and body fat percentage relative to position.
While there are many ways to judge accuracy, this particular model uses all players drafted in 2013 and prior to estimate what is and is not predictive to NBA success. Since those players were part of the “experiment” so to say, it is not fair to say we predicted them correctly (like taking a test with an answer sheet). Although those drafted 2014-’17 have not finished the first five years of their NBA careers, we can see how the model is doing so far and compare that to how scouts do.
So far, our model has done a better job at identifying top talent and has less noise. This is expected since we are using draft analysts as part of our model, using their rankings as an input. No one is a perfect talent evaluator, not even an analytics-based model, but by using analytics a team can increase the chance it selects an All-Star at the top of the draft and decrease the chance it lands a complete bust.
For more from ESPN Analytics, visit the ESPN Analytics Index.