Advanced Darlings
Who's Afraid of the DARK(O)?
I’m going to give you stat lines for two NBA players, and I want you to tell me which one you’d rather have on your team. Ready?
Hopefully that was a fairly easy exercise - why not take three times the points? Now, here’s how these players have been described by the media.
Zach Lowe on Player A: “The consensus — if there is one, and he’s got some fans, and he’s got some mega-detractors — but the consensus is kind of like empty calories ball hog”
Kendrick Perkins on Player B: He is “the key reason, outside of franchise guys, why [his team] is set to hang a banner this upcoming season.”
The blur wasn’t great, so you’ve likely already figured out that we’re talking about Cam Thomas and Alex Caruso. Thomas was just waived by his very bad Brooklyn Nets team, while Caruso is one of the most important players on the first-place Oklahoma City Thunder.
These two players epitomize opposite ends of the spectrum - Thomas puts up superficial numbers but doesn’t have a major impact on winning games, while Caruso does all the little things that don’t show up in the box score, but contributed to an NBA championship.1
“Advanced” stats try to capture this “impact on winning”. By comparing them to the box score numbers - which can be succinctly combined using Game Score2 - we can see which players diverge the most from their headline numbers. There’s an alphabet soup of advanced stats we could choose from - APM, BPM, DPM, EPM,3 LEBRON, RAPM, RAPTOR,4 VORP, WAR, WS to name a few - but for this exercise I’ll be using DARKO5 because 1) like Game Score it is also based on the box score, but heavily adjusted to capture the nebulous “impact” and 2) it’s readily available courtesy of CraftedNBA.
Plotting every player6 in the NBA along these two stats, we can see which players are the “empty calories ball hogs” and which are the “advanced darlings”. Game Score is on the x-axis with DARKO on the y, so the farther right a player is the more they fill the box score and the farther up the better all-around player they are. The dotted, 45-degree line is where the counting stats and advanced stats line up.
Thomas and Caruso are about where you’d expect them to be. Up in the blue area, where the impact > the box score, we find Caruso and other tenacious defenders in the Oklahoma City Caruso School of Defense: his teammates Kenrich Williams, Cason Wallace, and Aaron Wiggins.7 The Pelicans’ Herb Jones and the Magic’s Jalen Suggs are nearby as well.
DARKO also clearly loves big defenders who pull down tons of offensive rebounds. Draymond Green, Steven Adams, Mitchell Robinson, Rudy Gobert, Jarrett Allen, Bam Adebayo, and Chet Holmgren all rank in the top 10% of the league.
On the other side, along with Cam Thomas, guys like Paolo Banchero (averaging 21 points, 8 rebounds, and 5 assists per game), Josh Giddey (19, 9, and 9), and Trey Murphy (22, 6, and 4) are all having productive individual seasons, but their teams are significantly underperforming expectations.
And then there’s Russell Westbrook, the former MVP and king of the triple-doubles who continues to put up solid numbers … for the woeful Sacramento Kings. He’s actually not alone down there; most of the 12-44 Kangz roster is in the deep right corner of this 2x2, signifying numbers without wins.
More interestingly, many of the Nuggets players are also below the 45-degree line. All-Stars Nikola Jokić and Jamal Murray are among the best in the league by whatever stat you look at, but then there is a huge DARKO gap8 to recent two-way convert Spencer Jones. Peyton Watson, Tim Hardaway Jr., Jonas Valančiūnas, and Jalen Pickett have all been playing well for the Nuggets, keeping the team afloat and in the top three in the Western Conference while Jokić was out with injury, but their advanced stats don’t reflect that.
One explanation for the disconnect is that DARKO takes a long-term view on a player. While it does update daily and puts more weight on recent games, it considers every NBA game a player has ever played and can take a bit of time to reflect improvements. Guys like Watson, Pickett, and Strawther have been in the league for several years, but hadn’t been particularly productive until this recent stretch.9 Meanwhile, Spencer Jones had only played 20 NBA games prior to this season, so he’s not dragged down as much by past performance.
Jones has actually been one of the more productive young guys, as seen in his placement relative to this year’s crop of Rookies.
A lot of these Rookies are already above average in their box score contributions, and some (VJ Edgecombe on the Sixers, Dylan Harper on the Spurs) are even contributing to good teams! And then there’s Kon Knueppel, the fourth overall pick who is having one of the best rookie seasons ever, according to the advanced stats.
Knueppel10 is averaging 19 points, 6 rebounds, and 4 assists, while nailing 43% of his three-pointers, on pace for the seventh best rookie percentage ever while breaking the rookie record for made threes. He’s been a big part of the unlock in Charlotte that has the Hornets projected to make the Playoffs for the first time in the LaMelo Ball era.11
Cooper Flagg is not yet as beloved by the advanced stats. He slightly edges Knueppel in the box score, though as graphsketball pointed out in a very slick video, the Rookie of the Year race between Flagg and Knueppel is extremely close by the main counting stats.
Yet DARKO views Flagg as a typical starter on offense, and a below average defender. Most analysts still think Flagg will be the better NBA player in the long-run, which is a reminder that the advanced stats are just one lens through which we can evaluate players, but they don’t give the full picture. Anyway, it may be a fool’s errand to use overly complicated math to deny what Flagg does on the court.12
I realize Cam Thomas, playing with his new team in Milwaukee, just beat the Thunder and Caruso. The bigger picture still holds.
Game Score combines every column of a box score as follows: Points + 0.4 * Field Goals Made - 0.7 * Field Goals Attempted - 0.4*(Free Throws Attempted - Free Throws Made) + 0.7 * Offensive Rebounds + 0.3 * Defensive Rebounds + Steals + 0.7 * Assists + 0.7 * Blocks - 0.4 * Personal Fouls - Turnovers.
Looks like we need a CPM. I’ll start working on Charting hoops Plus Minus.
Minimum 30 games played. Data as of February 8.
Basically every Thunder player has a better DARKO rating than Game Score rating, and they all have extremely positive net ratings.
Aaron Gordon and Christian Braun haven’t played enough games to make this chart. Get well soon!
This doesn’t help explain the veteran Kings - they’ve just been bad.
For some reason this name is actually so much easier to spell than it looks like it should be.
It should be noted that DARKO actually really likes LaMelo as well, ranking him in the 91% percentile.
Does this final sentence invalidate the whole post? Maybe, but I still found it interesting, and I hope you did too!