The 2018 NBA Draft Player Comparisons
An easy way to think about NBA draft prospects is to compare them against players who we think they are most like. DeAndre Ayton is the next Shawn Kemp, Markelle Fultz’s best comp is James Harden (yes…), Josh Jackson is a jacked up Khris Middleton, the list goes on.
Computers and machine learning are being used projections of all kinds, so why not this? I’m a ML rookie so what I wanted to do was twofold — learn about maching learning and in the process see how a ML network would make draft comparisons. The idea was to use a simple neural network to help project a 2018 draft prospect’s impact and success in the NBA based on a known quantity — NBA players drafted before them — instead of comparing prospects to each other (unknown quantity since they have not played in league yet). Instead of projecting points / rebounds / steals / or other stats from college to the NBA, map prospects to former draftees and use the NBA player’s success as a projection. The result is that for each draft prospect a profile is created that shows what mix of former NCAA players they most resemble in college. Maybe it holds that if a draftee resembles players that turned NCAA success into NBA success, their chances of similar NBA success is greater.
To do this I pulled the game by game stat line for each college game for all the NCAA players drafted between 2011 and 2017. A model was built that created a profile for each of the 2011–2017 drafted players based on the stats in each of the college games they played, height, weight, position, etc. Then all of the career games of the top 100 prospects from Aran Smith’s Big Board were fed into the model. The result was each prospect game was mapped to a profile of a former draftee. The numbers of matches were summed up and plotted on pie charts. For example, Jaren Jackson’s chart is below:
Out of Jaren Jackson Jr’s 35 career games 18 mapped to Miles Turner, 16 to Bobby Portis, and 1 to Erik Murphy. If only one game mapped to a single former draftee (Erik Murphy in this case), the single game was discarded as not relevant to the overall picture and not plotted.
The reason for the 2011 cutoff was that that was the earliest game by game stats I could find. The data set used included over 21,000 games by over 300 players.
To verify the model and approach, it was also applied to players in the 2016 and 2017 drafts. The 2016 / 2017 analysis can be verified by seeing if the projections made for these players are reasonable as these players now have an NBA track record. If the 2016 / 2017 projections hold water, it is a validation of the model. Some of the highlights from running players drafted in 2016 and 2017:
In general, this is not meant to be a single measure of draftee success, but is interesting as a data point to be used in conjunction with other player analysis if valuation is your goal.
What this does not include:
· Players that didn’t play NCAA basketball. Foreign players were not included as the data set only included NCAA stats
· It does not factor in athletic ability. For example, while DeAndre Ayton was statistically similar to Tyler Zeller and Kyle Anderson, his athletic ability must be considered as a differentiator
Based on the results that were found, below are the players I’ll be keeping an eye on next season as they project to current NBA players who have found success:
Jaren Jackson Jr: Myles Turner and Bobby Portis
Mikal Bridges: Will Barton
Kevin Knox: Chandler Parsons (when healthy)
Shai Gilgeous-Alexander: Zach Lavine, Jeremy Lamb, Michael Carter-Williams (had a run)
Troy Brown: Taurean Prince, Khris Middleton
Trevon Duval: De’Aaron Fox, Brandon Knight
Jerome Robinson: Will Barton
Sviatoslav Mykhailiuk: Caris LeVert, Brandon Ingram, Tony Snell, Kelly Oubre, others
Jevon Carter: Damian Lillard
Jacob Evans: Chandler Parsons, Klay Thompson, Joel Embid, others
Kris Wilkes (withdrew from draft): Brandon Ingram, Andrew Wiggins
Malik Newman: C.J. McCollum, Kadeem Allen, Monte Morris, Joe Harris,
Chimezie Metu: Jonathan Isaac, Ivan Rabb, John Collins, John Henson, Alex Len
Jalen McDaniels (withdrew from draft): Kyle Kuzma
Justin Jackson (Maryland): Jaylen Brown, Kawhi Leonard, Jeff Taylor, others
You can browser search for players you are interested in or take some time to scroll through the charts for the complete list of players that qualified: