I wanted to put into perspective how deep this draft class is and put in perspective the type of season each top prospect had relative to an NBA player's college season(s). So, I decided to compute cosine similarity scores between the top prospects in this draft and NBA players’ NCAA seasons.
Dating back to the 1993 Draft, I took every top 5 pick, every all-star or all-NBA player drafted, and as the drafts got closer to the modern era, I included more players drafted who may not be all stars but are currently contributing on NBA teams (Marcus Smart, Andre Roberson, etc.), as well as non-Top 5 drafted players who may not be currently contributing but had some interesting NCAA seasons (Trey Burke, James Young, etc.). I ignored international and high school drafted players since they didn't have any NCAA data (sorry Dirk, Kobe, LeBron, etc.). I wanted to compare NCAA season to NCAA season.
I computed two cosine similarity values between all of those seasons and each prospect’s season. One was for their per 40 minutes stats and the other was for their advanced stats. Overall, I inputted 446 NCAA seasons from NBA players. I used sports-reference.com to get this data. However, the 1997-98 NCAA per 40 minutes stats were incomplete, as well as every player’s NCAA seasons from the 2009 Draft, which was really odd. So no Griffin, Curry, or Harden comps. Additionally, I only computed the advanced stat similarity on seasons from 2010-11 onward, as every season beforehand was incomplete in someway. So, I returned two sets of the 30 most similar seasonal per 40 minute stat lines and the 30 most similar seasonal advanced stat lines. I then took the intersection of both sets to get the “most similar” seasons produced by an NBA player in the NCAA for each prospect.
Here’s the github repo for all the nerds. I separated the results by guards, wings, and bigs. For the per 40 min. similarity, I took into account FG%, 2P%, 3P%, FT%, PTS, AST, TRB, STL, and BLKs. For the adv stats similarity, I took into account TS%, eFG%, FTr, WS/40, ORB%, DRB%, AST%, STL%, BLK%, TOV%, USG%, PER, OBPM, and DBPM. The results were interesting. Some results made a lot of sense, others… not as much. I came here to share the results and write a little about the consensus top 10 freshmen, and then write a little less about some of the other prospects in the dataset. So, here we go:
Some terms I will be using:
Unique p40 == an NBA player’s NCAA season per 40 minutes stat line that was only returned for this prospect.
Unique adv == an NBA player’s NCAA season advanced stat line that was only returned for this prospect.
(xNumber) means the number of NCAA seasons from that particular NBA player.
F., So.,J.,Se. == Freshman, Sophomore, Junior, and Senior years, respectively.
The order in which I wrote the first 10 players is my opinion based on how transcendent they could be. Determining an order for 3-7 was quite challenging. Don't hate on me for being high or low on somebody because I think the first 10 players are all really good prospects.
1. Lonzo Ball - The most unique prospect in the dataset. His stats were so ridiculous that he returned the lowest similarity scores in the system. Jason Kidd’s sophomore year was Lonzo’s #1 per 40 minute similarity score. He was the only player to return a single season of Jason Kidd, and he managed to return both Kidd’s freshmen year and sophomore year. He also returned both of Rondo’s Kentucky seasons, which shows what Ball can be if his jump shot doesn’t translate. The system also returned both of MCW’s Syracuse seasons, two Kris Dunn seasons, and Evan Turner’s magical Junior OSU season, showing that he can bust just as easily as he can turn into the next JKidd. However, he produced unprecedented stats for a Freshman guard and appears to have a legendary BBIQ, which is why I give him the slight edge over Fultz.
Unique p40: Kidd (x2), Conley.
Others returned: Westbrook, Grant Hill, Artest, Penny (x2), Deron Williams (x2), Bibby, Wall, DRose.
2. Markelle Fultz - The only player in the dataset to return both of Westbrook’s UCLA seasons. Additionally, him and Fox were the only players to return an Iverson season. Fultz, however, returned the more impressive AI season (Sophomore). I was a little surprised that Penny didn’t show up given how similar his Junior season was. Fultz is a ferocious scorer, and he’s going to get a lot of buckets in his career. Kyrie Irving showed up in the intersection set, but so did Reggie Jackson, Trey Burke, and Dion Waiters. His scoring potential is outrageous though, exemplified by Westbrook and Iverson showing up in the per40 set, which is why he’ll be the #1 pick in the draft, presumably.
Unique p40: Westbrook (x2), Iverson (So.), Baron Davis, Avery Bradley.
Others returned: Francis, Marbury, Arenas, Irving, Roy, Rose, DRussell.
3. De’Aaron Fox - The only top guard to return a Dwyane Wade season, which is Fox’s absolute ceiling if he never develops a 3-point shot. Wade shot 28.7% from 3 for his career and racked up a 57.8 career VORP. They actually resemble each other in that they’re not good 3-point shooters, they shoot in the mid-70s from the line, they’re great defenders, and they’re ridiculous slashers. He also was the only prospect to return AI’s freshmen season (in which he shot 23.2% from 3 on 151 attempts). Fox is crazy fast and quick, so these two comps fit his absolute ceiling. A more realistic pessimistic outcome, is Marcus Smart or Elfrid Payton, who showed up in the intersection set. However, if Fox develops a jumper or puts it together like DWade or AI did with a below average 3-point shot, he will be a franchise guard. I have confidence that his slashing ability, pull up game, defense, and overall athleticism will allow him to prosper on a court with more space than in college. I also have confidence that his jumper will improve with reps because he probably never had to shoot in games due to him being quicker than everybody and his form is pretty damn crispy (maybe it just looks that way because he's a lefty). If he becomes a 3-point threat, it’s game over.
Unique p40: Iverson (F.).
Others returned: Wade, Westbrook, Francis, Penny, Marbury, Arenas, Wall, Lillard, Kemba, Roy, IT4 (x2), DRose.
4. Jonathan Isaac - Isaac may have returned the most accurate results based on the eye test. He might be the best overall defender in the draft (maybe a little less perimeter defense than Jackson, but more rim protection), and the results show this. He was the only prospect to return two seasons of Kawhi. His advanced stats similarity set returned all three of Roberson’s Colorado seasons, when no other 1st round prospect returned one. His #1 per 40 min. comp returned was Al-Farouq Aminu, which exemplifies the role he could fill in the NBA by being able to play the 3, 4, or even small ball 5. Behind Ball, he might be the most unique guy in the class. While he might be too passive or too much of a tweener to be a superstar, he could be a super role player with all-NBA defense like what Iguodala is on the Warriors. However, I think his ceiling is absolutely outrageous just because of his combination of defense, length, and shooting. Also interestingly, Isaac was the only guard or wing to return Anthony Davis and Karl Towns in the advanced stats set.
Unique p40: Kawhi (x2).
Unique adv: Roberson (x3).
Others returned: Carter, Marion, Bosh, Gerald Wallace, Gordon Hayward.
5. Malik Monk - No one returned a single JJ Redick Duke season; Monk returned all 4. Monk also returned Kyrie, Kemba, Rip Hamilton, and Isaiah Thomas seasons showing that he could be more than just a shooter at the NBA level. Unfortunately, defense is half the game, and he’ll have to improve his tenacity on that end. The system returned Devin Booker as the most similar player, (#4 in per40 behind Redick’s senior year, Afflalo, and Eric Gordon; #2 in adv to Gary Harris). Guys like OJ Mayo and Nik Stauskas were returned, showing Monk’s floor. However, if Monk shows that he was as underutilized as Booker was at Kentucky, then the sky’s the limit for his scoring potential.
Unique p40: Lillard (x3), Redick (x4), Eric Gordon.
Others returned: Irving, Kemba, IT4, Rip Hamilton, Klay (x2), McCollum (x2).
6. Josh Jackson - I had him at 3 until I embarked on this project. He wasn’t getting ranked above Ball and Fultz. Fox and Isaac returned more interesting results, and Monk could be the best scorer out of everyone not named Fultz. The system returned Justise Winslow as Jackson’s most similar player. He was #2 to Oladipo in the per40 set and #2 to Taurean Prince in the adv set. I asked myself if I would take Winslow over any of the ridiculously talented guards in this draft and that's a tough question. While I think Winslow is a great player and still has upside, he is limited to 3-D wing without the ‘3’ (at least for now). I’m afraid that this is the fate of Jackson. However, Jackson can do so many other things well, which is why I have him above DSJ. He might be the best perimeter defender in the draft, he rebounds and passes well for a wing, can slash easily, and finish with authority. If he figures out his jumper, he could wind up being the best player in the draft. His shooting mechanics aren't as clean as Fox's, which is why I'm a little higher on Fox than Jackson.
Unique p40: Josh Howard (x3)
Others returned: Pierce, Carter, Artest, Jason Richardson, Granger, Roy, Stackhouse.
7. Dennis Smith Jr. - One of three prospects to return Chris Paul (along with Jawun Evans and Monte Morris). In both sets, the most similar season was identical. That was Isaiah Thomas’ final season at Washington. DSJ returned some impressive names, but I’m worried about his BBIQ and defense. The system returned Isaiah Thomas, Trey Burke, Tyler Ennis, and Kyrie Irving as the 4 most similar players. DSJ returned enough impressive guards to keep him in the top 7. Out of the prospects in my 3-7 range, I think Fox is the best player, think Isaac has the highest ceiling, think Monk is the best fit in the modern NBA, and Jackson has the most 2-way potential. I feel like Dennis Smith Jr. is the wildcard. He could turn out to be the best player in the group or bust entirely.
Unique p40: None
Others returned: CP3, Westbrook, Francis, Marbury, Bibby, Wall, Irving, Kemba, IT4, DRose.
8. Jayson Tatum - The system was kind to Tatum, returning some supreme talents. However, I think this ranking is where he fits, below five guards who will be impact players in an increasingly small-ball, modern NBA, and below two more versatile forwards. Tatum put up some impressive numbers at Duke, which is why Durant and Melo’s Freshmen seasons were returned. Also, Paul George makes his lone appearance here, showing what Tatum’s absolute ceiling is. However, I think a more reasonable comp is Tobias Harris, someone who showed up in the intersection set. Harris is a decent defender and a decent offensive player, and averaged 16/5/2 on a 48/35/84 shooting line this past season for the Pistons. I can see Tatum averaging that if he puts it all together, but I don't know if he's elite at anything, just like Harris. I'm also really glad Rudy Gay showed up here because I feel like Tatum is the second coming of him.
Unique p40: Paul George, Michael Redd.
Others returned: Durant, Carmelo, Jimmy Butler, Rudy Gay.
9. Lauri Markkanen - Lauri Markannen’s top 2 most similar per 40 minute stat lines were both of Ryan Anderson’s seasons at Cal. I’m pretty confident he can step in and reasonably fill a Ryan Anderson role on a team as a stretch 4/5. The reason why he’s this low is because his defense is pathetic. 0.7 blocks per 40 minutes is absolutely terrible for a 7 footer. However, he can clearly shoot, and 7 footers who can hit threes at an elite clip are certainly valuable commodities nowadays.
Unique p40: Van Horn (x4).
Others returned: Durant, Carmelo, Szczerbiak.
10. Zach Collins - The advanced stats God. Here are some of the players in the advanced stats similarity set for Collins: Anthony Davis, Karl Towns, Embiid, Drummond, Steven Adams, Tristan Thompson, Nerlens, Roberson, and Myles Turner. Collins produced a godly 30.9 PER and the highest Win Shares per 40 minutes out of any prospect I put into the system (there were 46 of them). His stats are a little inflated since he played in the WCC, but it’s still impressive considering he was a Freshman. The system returned Richaun Holmes as Collins' most similar player (Richaun's Junior and Sophomore years were the Top 2 most similar in the intersection set). I can see Collins filling a RichHolmieChaun role for a team: energy, shot blocking, and the occasional 3 pointer, with the ceiling to eventually be a starting center.
Unique p40: None
Others returned: Webber, Bosh, Love, Vucevic.
Some other prospects I included in the dataset but won't write about as much as I did the first 10:
Donovan Mitchell - Returned a Billups season, a Nash season, and a Rip Hamilton season. Most similar player was Gary Harris. He has the potential to be a really good 2-way guard, and he'd probably be 11th on my board. After Mitchell and Ntilikina, I have no clue how to rank anybody else.
Jawun Evans - Returned four Steve Nash seasons! Also, a CP3, Billups, Marbury, Bibby, Deron Williams, Wall, Kyrie, Kemba, and IT4 season. His top 3 most similar players were Tyler Ennis, Trey Burke, and Isaiah Thomas, exemplifying his low floor and high ceiling.
Luke Kennard - Returned a Sugar Ray season. Most similar player was fellow lefty shooting guard Dukie Rodney Hood.
Tyler Dorsey - Returned two Sugar Ray seasons. Interestingly, Jamal Murray’s Kentucky season was the #1 similar per 40 minute stat line and #1 advanced stat line.
Sindarious Thornwell - Also return two Sugar Ray seasons. Most similar player was CJ McCollum, showing his scoring potential. Thornwell had a really unique season, blending elite scoring with steals and blocks and could be great value in the 2nd round (if he doesn't get scooped in the late 1st).
Justin Jackson (UNC) - Allen Crabbe showed up in his intersection set. I feel like he could be Allen Crabbe.
Monte Morris - Only player to return both of Chris Paul’s Wake Forest seasons. Also returned all four of Jameer Nelson’s St. Joe’s seasons showing his capability of being a floor general. Most similar player was Trey Burke.
OG Anunoby - Only player to return both of CWebb’s Michigan seasons. I think a fitting comp for him is Jae Crowder, who was his fifth most similar player. Both of Oladipo's Indiana seasons also appeared in the intersection set.
Semi Ojeleye - Also returned two seasons of Sugar Ray. Most similar player was Doug McDermott. I guess that’s what happens when you shoot 3s at a high clip but only average 0.5 steals and 0.5 blocks per 40 minutes.
John Collins - The top 2 most similar seasons were Cody Zeller’s Freshman and Sophomore years at Indiana.
DJ Wilson - I think he has a ton of defensive promise having watched a lot of Michigan games. Most similar player was Bobby Portis
Nigel Hayes - The only Ben Simmons appearance in the dataset (returned in the advanced similarity set)! His three most similar players were Will Barton, Justise Winslow, and Marcus Morris, which I don't know how to interpret.
Jordan Bell - The only prospect to return a Tim Duncan season. His advanced stats similarity set included Draymond, Roberson, Nerlens, Embiid, Towns, Drummond, Adams, and Kyle Anderson, showing his defensive potential. His three most similar players were Otto Porter, Joel Embiid, and Alex Len, which is one of the weirdest results in the entire system.
Harry Giles - Not really sure how to interpret his Freshman season, but the top 3 most similar players were Steven Adams, Thomas Robinson, and Andre Drummond.
Caleb Swanigan - Most similar player was Draymond, sleeper alert? Also returned Lamar Odom, Kevin Love, and Chris Bosh seasons.
Bam Adebayo - Only one player was in his intersection set, and that was Tristan Thompson.
There are more prospects in the dataset if you want to check out the results. I really wanted to include LaVar Ball in the dataset, but his lone season was 1987-1988 which had no per 40 minutes stats on sports-reference.
