Nylon Calculus: What can we expect from Donovan Mitchell in year two?

He ultimately lost out on the Rookie of the Year Award to Ben Simmons, but Donovan Mitchell had a great rookie season any way you look at it. It was a year-long debate but Mitchell had a strong statistical resume, helped the Jazz make the playoffs and then continued to carry their offense once they hit the postseason.

Mitchell has taken some flak for a statistical resume that, at the surface, may not have been as eye-popping as Simmons’. However, he ranked above the 80th percentile among rookies in almost every advanced stat. Among all rookies since the 1979-80 season, 2710 distinct players, he ranks among some of the most impressive players across many of these categories

The chart below shows the highest percentile ranks for Mitchell’s rookie season stats compared with all rookie seasons in the last 40 years.

The most relevant metrics include VORP at almost 98th percentile, USG% at the 95th percentile and BPM at the 93rd percentile. Select almost any metric and Mitchell excelled at it when compared to all rookies since the 3-point line was implemented.

Not only was Mitchell excellent in the regular season, but he also performed at a high level in the playoffs. When compared with other rookies in the playoffs he is near the top. For a rookie in his first postseason, his scoring output is comparable to Kareem Abdul-Jabbar. He led the Utah Jazz to a first-round win over the Oklahoma City Thunder and he owns the Jazz rookie playoff record for points in a game. With Ricky Rubio out due to injury, Mitchell also led the Jazz in points, assists and steals against the Houston Rockets.

READ MORE: Did the Warriors really ruin the NBA?

Mitchell had a historically great rookie regular season and playoffs. He also won the 2018 dunk contest with a solid Vince Carter dunk impression. With such great first season, what can we expect from him next season?

I started by looking at historical rookie data in order to get an idea of similar players. I gathered data on all rookies since the 1979-80 season from Basketball-Reference.com. With advanced and per-36 stats on over 2700 rookies, I then attempted to find the most comparable rookies in the last 40 years. I used a nearest neighbors algorithm in order to find the closest comparisons on over 40 different statistics. The goal with this algorithm is to find the players whose stats are closest to each other in terms of standard Euclidean distance. If you remember your high school geometry class the process is very similar to the distance formula. Instead of only two dimensions, it uses over 40 different stats to calculate the points that are closest to one another.

Based on the above nearest neighbors algorithm, Donovan Mitchell’s 10 nearest neighbors are listed below.

These are some promising comparisons with two Hall of Fame players in Mourning and Kidd as well as three All-NBA comparisons in Westbrook, Hardaway and Towns. One drawback to this method is that bigs like Mourning and Towns are statistically similar to a player like Mitchell. Towns and Mourning excel in blocks and rebounds but lack the shooting ability of Mitchell. While these players play a completely different type of game the rest of their stats are very similar. VORP, BPM, the various WS stats and more basic stats such as PTS and eFG% are all very close. With over 40 stats used in the comparison, the difference between the handful of rebounding and shooting stats is minimized.

To handle some of the discrepancies with the style of play I tried adding weights to the various stats with little change in the outcome. Other steps to possibly improve these comparisons would be to include other stats such as tracking or shot type. Unfortunately, the tracking and shot type stats only started being tracked in the 2013-14 season and I wanted a more historical look.

After finding Mitchell’s nearest neighbors I then moved on to predicting his performance for the 2018-19 season. In order to do this, I averaged the difference from each of the neighbors’ rookie seasons to their second seasons. I then added this average difference to Mitchell’s rookie season stats to get his predictions. I also added in a 90 percent confidence interval to give a range of likely production for each stat. The images below show the projected stats as well as the likely range.

For the most part, we can expect more of the same from Mitchell next season. Projected points per 36 minutes are over 23 and the model expects an increase of VORP to 3.02 and BPM of 2.1. We can expect Mitchell to shoot even more 3s with an increase in 3PAr and also a higher AST% at nearly 22 percent. Some drawbacks could be a lower 3PT%, lower FTr and slightly higher TOV per 36.

If you would like to see this data in more detail or look at other 2017-18 rookies check out the full app here. It’s a simple web app that was the basis for this post. The nearest neighbors algorithm has its benefits as it gives decently accurate comparable players using a standard distance calculation. This process for calculating projections helps maintain simplicity and ease of understanding. In order to improve the accuracy, we could use a more complex algorithm such as K Means or spectral clustering. Another option might include adding actual labels to each player such as Hall of Fame, All-NBA, All-Star, starter, bench player, etc. Once each player is labeled we could use a more complex algorithm including ensemble or SVC classifiers to choose similar players. I chose the nearest neighbors for the main reasons of simplicity and ease of understanding of the process.

Looking at the projected stats we see a fairly large range of possibilities. We can expect with 90 percent certainty that his production will fall somewhere in the range for each of these stats. For example, barring injury or other unforeseen events, we can expect Mitchell to average between 19.4 and 26.7 pooints per 36 minutes. However, I think it is much more likely he performs at the upper end of these ranges.

There are a few reasons for a positive outlook for Mitchell. First, he plays for a quality organization with a history of sustained success including back to back Finals trips and 20 straight playoff seasons from 1983-2002.  Also, Quin Snyder is a well-respected coach and was a finalist for Coach of the Year in 2017-18.  Secondly, Mitchell is dedicated to becoming a better player, which is not true for a lot of rookies. Finally, the one stat that is projected to decline is his 3PT%, from almost 35 percent down to 33 percent. He makes up for the drop in percentage with a higher 3PAr, which means he might be less efficient next year. I think the drop in percentage is less likely for Mitchell due to the added emphasis on the 3-pointer throughout the league. A lot of Mitchell’s comps were from players in an era that was not as focused on 3s, and improving 3PT% is now a common goal throughout the league.

One season makes it difficult to project accurately into the future but Donovan Mitchell’s first season has been historically impressive. If he maintains his drive to excel and stays healthy, his future is very bright.