Re-introducing an NBA Statistical Player Comparison Tool

Nov 15, 2015; Minneapolis, MN, USA; Minnesota Timberwolves center Karl-Anthony Towns (32) dunks the ball in the fourth quarter against the Memphis Grizzlies at Target Center. The Grizzlies won 114-106. Mandatory Credit: Brad Rempel-USA TODAY Sports
Nov 15, 2015; Minneapolis, MN, USA; Minnesota Timberwolves center Karl-Anthony Towns (32) dunks the ball in the fourth quarter against the Memphis Grizzlies at Target Center. The Grizzlies won 114-106. Mandatory Credit: Brad Rempel-USA TODAY Sports /
facebooktwitterreddit
Nov 15, 2015; Minneapolis, MN, USA; Minnesota Timberwolves center Karl-Anthony Towns (32) dunks the ball in the fourth quarter against the Memphis Grizzlies at Target Center. The Grizzlies won 114-106. Mandatory Credit: Brad Rempel-USA TODAY Sports
Nov 15, 2015; Minneapolis, MN, USA; Minnesota Timberwolves center Karl-Anthony Towns (32) dunks the ball in the fourth quarter against the Memphis Grizzlies at Target Center. The Grizzlies won 114-106. Mandatory Credit: Brad Rempel-USA TODAY Sports /

Basketball fans love coming up with comparisons for young NBA players.  These comparisons act as heuristics for understanding how a player might fit into different lineups and projecting players forward in their NBA careers.  The problem is that brainstorming comparisons is difficult.  Far too often, fans fall back on the same archetypical players (e.g. Kyle Korver, Reggie Evans, Kobe..), or are limited to a stunted time depth.  It is simply too difficult for even expert fans to call up less obvious comparable players from the annals of NBA history.

My recently updated player-season comparison tool is designed to aid in this process.  This tool looks at the boxscore production, basic physical profile, and age of NBA players and compares these data to past NBA seasons.  This results in a list of the most similar player-seasons of the past.  This list is an objectively accurate description of which players put up the closest numbers, but there are many things the comparison tool’s outputs are not:

  1. Context informed: These comparisons take no account of team, or even league context. The model simply looks at the production rate of different boxscore statistics. It does not care how players got to those numbers or what their role was on their team.
  2. Destiny: It is easy to say “well my team’s prospect is performing similar to this guy who was never any good, so we should probably give up on him.” Statistical production in a given season, especially for young players, only says so much about their future. Other indicators like prior hype (captured in things like draft slot), work-ethic, cognitive ability, and athleticism can put similar-producing players on very different expected developmental curves.
  3. Precise measure of value: The comparison tool will sometimes identify a close statistical comparison where the actual value of the two players is wildly divergent. This happens because one player might be similar to, but worse, across most measures, or critically limited in a specific area.  If this bothers you, feel free to crank up the slider for the weighting on “Win Shares.”  Also, be sure to pay attention to the “Val” column.  Positive scores in the column indicate “rich man’s version” while negative scores indicate “poor man’s version”.  Value is in standardized units, with anything over 1 being considered a very large difference.
  4. Stylistically valid: Good comparisons are ultimately going to take into account details of the game the boxscore completely ignores. This leads to occasional comps that might look great on paper, but are immediately hilarious to anyone who has actually watched the two players (Blake Griffin drawing Dan Issel for example).
  5. Consistently accurate across comparisons: Some players are easy to find comparisons for, while others are not. There are cases where one guy has 50 past player-seasons more similar to his than another player is to his closest comparison. Pay attention to the “SD” column, which indicates just how close a comparison is.  As a rule of thumb, anything in the 0.2s is very close, the 0.3s are still pretty good comparisons, then once you get past 0.5s you really should not take any comparisons too seriously.

To help hammer home how to interpret player-comparison outputs, I am going to take you through a few examples.  Let’s start with the rookies drafted in the top 5:

Karl Towns:

twonscomp
twonscomp /

Notice the Val column.  Brook Lopez’ first NBA season is the closest comparison to Towns’ production so far, but Towns has been much better overall than Brook was as a rookie.  I also recommend going to the bottom of the sliders list and removing the age component.  You basically get a long list of Tim Duncan seasons (FWIW, Towns’ NCAA productions was also very similar to Duncan’s work at Wake Forest)

D’Angelo Russell:

dangelocoms
dangelocoms /

Russell has had a rough time of things so far, but that is not uncommon for young lead guards.  Despite his struggles, D’Angelo still draws a relatively strong list of comparable players.

Jahlil Okafor:

jahliloka
jahliloka /

Again, note the Val column.  Unlike with Towns, Okafor’s relative scores are going in the wrong direction.  His overall boxscore looks similar to past all-offense young bigs Shareef Abdur-Rahim and Eddy Curry, but Okafor’s inability to get to the line leaves him a solid tier or two below them in value.

Kristaps Porzingis:

porzicomp
porzicomp /

Note that Kristaps is much taller than any of the players who put up similar production at the same age.  This makes him a difficult comparison, with no player-season drawing an “SD” value below 0.4 (the same applies to Okafor by the way).  In addition, this list really captures the positional ambiguity with Porzingis.  His statistical footprint is that of an NBA power-forward, but his height makes him an obvious candidate to play the 5 longterm.

Mario Hezonja:

hezonjacomps
hezonjacomps /

Hezonja is throwing up three-point shots at a very high rate so far in his rookie season.  Unsurprisingly, this leads to a list of players who did little beyond shooting early in their careers.  Stauskas’ recent disappointing rookie season rates as the most similar outing, which probably is not something Magic fans want to see.  Hezonja’s physical profile and shooting ability are pretty special, so let’s hope he figures out what he is capable of as he starts to get comfortable in the NBA.