We live in the golden age of NBA data. From shot charts to adjusted on/off splits to advanced player tracking, the statistical toolkit available to the average fan today would make a 2003 front office exec blush. And that's a good thing. More information usually is.
But there's a problem: with enough stats, you can prove anything.
Want to argue that Player X is a top-5 defender at his position? There's a stat for that. Want to show that he's a complete liability who should be benched? There's a stat for that, too. The explosion of data has democratized analysis—but it's also flooded the field with narrative cherry-picking.
The Luka Experiment
Let's test this. I'm going to make a case I know is false:
Luka Doncic is a plus defender.
Before we compare him to those who play on ball across the league, let us look at how he stacks up against his own teammates. When looking at Doncic's stats aggregated across his stints with both the Lakers and the Mavs, he ranks third among all players on both teams in defensive impact.
His teams' defensive rating when he was on the floor was more than five and a half points per 100 possessions better than when he was off. The only players ahead of him in on/off defensive efficiency were Dereck Lively and Dorian Finney-Smith.
The Steal Rate Story
Okay, you're not a fan of on/off metrics? Let us look at his more granular stats. Luka "Quickest Hands in the West" Doncic had the highest steal rate among any player on either team. Not just higher than cones like Austin Reaves or our old GOAT LeBron James—higher than anyone. And, no, I'm not forgetting about 5x All-Defense selection Anthony Davis.
Rim Protection Numbers
And what about rim deterrence? According to Cleaning The Glass, Doncic limits opponent rim frequency better than nearly anyone he played with this year. Only Davis and Lively did it better.
Side note: should we be giving Lively more respect, or do we somehow still undervalue bigger defenders? I digress.
Expand the zone to include short midrange shots as well as rim shots, and Doncic ranks second at limiting all interior attempts. The only person ahead of him? You guessed it: Dereck Lively.
League-Wide Comparisons
So, is Doncic the best defender in LA or Dallas? If not, you must concede that he's in the top three. Still not convinced? Let's stop dunking on his teammates and look at how Doncic compares to other guards.
Among guards and combo guards, which ends up being the most apt comparison group for Doncic by Cleaning the Glass's standards, Doncic ranks:
- 5th in on/off defensive rating
- 9th in steal rate
- Above average in block rate
What more could you want from him? Put differently: by all the stats commonly used in defensive player evaluation, Luka Doncic is an elite defensive guard.
Point Guard Excellence
These stats become even more remarkable if you narrow the criteria to only those who CTG calls point guards. Among his positional contemporaries, Doncic finished the year:
- 3rd in steal rate
- Just outside the top ten in block rate
- FIRST in on/off defensive rating!
Move over Jalen Suggs, forget about Davion Mitchell, there's a new sheriff in town.
The Reality Check
And yet. And yet, if you've ever watched him play defense for five minutes, or if you watched the 2024 NBA finals, you know this cannot possibly be true.
So what happened?
The Framing Problem
A lot of this comes down to framing.
Handpicking stats to fit a certain narrative is far from the only way that analysts can get the numbers to work in their favor. Another way that I see this happen is through the use of percentiles. Specifically positional percentiles.
Switch the raw numbers to percentiles, and a player can go from average to impressive. Change his positional label—say, from forward to combo guard—and suddenly he ranks among the elite in block rate or defensive usage.
The Thompson Twins Case Study
Take two players with similar builds and play styles, but who are classified differently. Maybe one is listed as a wing, the other a guard. Their percentiles across stats like steal rate, contest frequency, and isolation points per possession can diverge dramatically, even if their raw stats are almost identical.
Example: The NBA's most electric twin experiment: Amen and Ausar Thompson.
According to Cleaning the Glass:
- Amen: Classified as a combo guard
- Ausar: Classified as a wing
Same (exact) build. Same chaos-inducing defense. Same downhill athleticism that breaks rotations before they're even set. But their stat profiles? Night and day.
- Ausar has slightly higher usage (less than a percentage point) but ranks 39 percentile points higher
- Amen has a higher assist rate than Ausar, and still gets buried 35 percentile points lower
The numbers didn't change. The reference group did. And suddenly the story shifts—not because of performance, but because of who a player is being compared to.
The Ubiquity of Positional Percentiles
But who even looks at positional percentiles you ask? Everyone. We all love Cleaning the Glass, and rightfully so. This is widely considered the gold standard of NBA stats. Whether you're looking at a graphic or reading an article, percentile stats are almost always referenced from CTG.
The problem: CTG only shows percentiles grouped by position. Unless analysts or twitter admins are calculating their own percentiles (they're not), you're seeing positional percentiles.
This matters. A lot.
So Should We Ignore Analytics?
Nope. You should definitely be paying attention to stats. Most of the time, advanced stats paint us a valuable picture. Plus, who is any analyst to say that a stat is lying? Numbers are objective, people are not.
However.
If the data can make Luka Doncic look like an elite defender, it can make anyone look like anything. And that's not a knock on Luka—he's a generational offensive player. It's a warning to you, the analyst, scout, or fan.
The Bottom Line
So yes, embrace the stats. Use them. Dive into them.
But do your own research. Don't judge a player off a set of handpicked stats. And certainly don't stop watching the games.