The Marginal Possession ≠ the Average PossessionThe Marginal Possession != the Average Possession

I came across a tweet from the ALL NBA Podcast (@ALLCITY_NBA) that said, “A possession with a paint touch is worth ~0.26 points per possession more on average than a possession without one.”

It is, as they stated, a random fact. Still, whenever I see things like this, my first instinct is to think about what I might do differently given this information.

A naive interpretation of this stat might be something along the lines of “Hey, let’s design an offense that optimizes for getting a paint touch. The data says if we convert 10 non-paint-touch possessions to paint-touch possessions, that’s an extra 2.6 points per game!”

If someone makes the case that the Magic should get into the paint more frequently, they need to be able to identify the specific possessions that would be better converted to paint-touch possessions and why a change in approach would improve them.

2016–2025 values are year-end adjusted closes; 2026 is the September 4 close. Sources: Macrotrends and FinanceCharts. CEO dates: Nike 2023 Form 10-K and Nike fiscal 2025 first-quarter release. Price movement alone does not establish the effect of a management decision.

The “paint touch advantage” might actually precede the touching of the paint

Imagine yourself with me in the Kia Center. Section 105, row 15. Close enough to pick up on subtle player positioning but high enough up to see the whole court.

Suggs drives right off of a Wendell Carter Jr. screen, forcing Paolo’s defender to shade towards the elbow. Suggs kicks it out to Paolo on the wing, who then fires the ball back to a rolling Carter Jr. in the middle of the paint. Carter Jr. is met by a crashing help-side defender, cutting off his lane to the rim. Carter Jr. makes one more pass to the opposite corner, where Bane is open for a high-percentage 3-point attempt.

The paint touch helped to create the open 3, but so did a number of other factors along the way. The definition of a “paint touch” is precise because it is tied to a specific area on the court. What the paint touch represents is not restricted to the 304 square feet between the free-throw line and the baseline. The action that allowed Paolo to find Carter Jr. in the lane, collapsing the defense and leaving Bane open in the corner, is the real advantage-creating play.

Now imagine the same action, except the defenders in the pick and roll (PnR) switch. When Suggs kicks it to Paolo, his defender is set, disallowing a drive. Without the defense having to rotate and close out, Carter Jr. catches the entry pass with nowhere to go and no help-side defensive crash. Bane’s defender is able to stay in the passing lane and the possession is dead in the water. The possession still includes a paint touch, but it does not create the same opportunity.

Another thing to think about: if a paint-entry attempt ends in a turnover before the paint is reached, this possession falls into the “no touch” group. Both paint entries and failed paint-entry attempts have the same intended outcome (paint touch) but are classified in opposing buckets. The implication here is that the two groups don’t cleanly separate possessions that tried to reach the paint from those that didn’t. Only paint-entry attempts that reach the paint show up in the paint-touch group.

Possession path	Intended paint attack	Observed paint touch
The entry reaches the paint	Yes	Yes
A turnover ends the entry first	Yes	No
The offense does not attack the paint	No	No

Definition: NBA Stats defines a paint touch as receiving the ball inside the three-second lane. The rows illustrate the classification problem described in the text; they are not observed NBA counts.

The 0.26 gap is not very instructive now that we’ve walked through the above. In order to evaluate the outcome of an attempt to attack the paint, we need to consider successful entries, turnovers, kickouts, and passed-up shots. Then we need to compare those results with what the offense could have generated from the same situations without that intent.

The diagram uses only the +0.26 value quoted in the linked ALLCITY_NBA post. The marginal comparison is intentionally unknown.

The Efficient Market Possession Hypothesis

Do teams make suboptimal decisions around which shots to take and which to pass up during a possession from time to time? No. Kidding. Yes. Of course they do. But probably not often. I’m not going to spend any time trying to verify this. If presented with an open dunk in the waning minutes of a crucial playoff game, an NBA player is going to take that dunk 10 out of 10 times, right?

The point still stands, though. If there’s an open lane and two easy points to be had, more often than not, a team is going to capitalize. So, if we expect all of those open paint touches to have been taken, we should then expect that the marginal paint touch will carry with it a lower expected outcome. It’s just not as easy as getting more paint touches = more points.

We’ve crunched the numbers (we have not, indeed, crunched the numbers, but I’ve shot a basketball from different places on the court and am comfortable relying on my empirical evidence), and the data are clear: ceteris paribus, it’s easier to make a shot taken closer to the rim than one taken from further away. We just need to take care of that nasty, inconvenient ceteris paribus. That’s where good coaching and player instinct come in! How do we generate additional paint touches that do not lower the average expected value of the paint-touch possession?

But wait, can’t we get a little worse if we do a lot more?

Ok, so maybe I was being a little simplistic and rigid above. Do we really need to make sure that, if we generate a marginal paint touch, the resulting attempt is no more difficult than the baseline? Not quite.

Crossing over into my professional world, think about users signing up for your product. If you are converting 10% of site visitors to customers, then 100 visitors will yield 10 customers. We could do things to “open up the funnel” and bring more visitors to the site, but these new visitors may have inherently weaker intent (we had to do something additional to get them there, anyway). So maybe if we open up that funnel to get 200 users, the conversion rate drops down to 8%. That’s OK because we now end up with 16 customers instead of 10. Back to basketball. The problem to solve here is: How do we generate more offensive opportunities? That could mean playing faster, forcing more turnovers, or extending possessions with offensive rebounds.

Creating more opportunities is one route. Improving the allocation of the opportunities already available is another.

Could we steal from bad possessions and give some to the paint?

Another scenario in which we can come out ahead by reducing the average points per paint touch is by turning a heavily contested long 2 into a heavily contested short 2. That heavily contested short 2 may carry an expected value (EV) lower than the average from that spot, but if it’s greater than the EV of a different shot, it’s net positive. Often, the clock can be the proverbial foot on the neck of the possession, applying more pressure as the seconds tick down. You can’t be picky with 2 seconds left on the shot clock. Brian Skinner formalizes the end-of-clock tradeoff here. It’s the basketball equivalent of someone’s standards going down at the bar as the clock marches closer to closing time.

Model element	Implication
Time and expected opportunities remaining	The acceptable shot-quality threshold falls as options run out.
Risk of losing the ball before another shot	Turnover risk makes earlier shot opportunities more valuable.
Distribution of future opportunities	The result depends on assumptions about chances that cannot be observed when passed up.

Source: Brian Skinner, “The Problem of Shot Selection in Basketball,” PLOS ONE 7(1), 2012.

Ok, I’m following so far but can you create an analogy using an obscure economic policy paper to really throw me?

Reader, you are in luck. So far, we’ve been focused on what’s going on with the offense in these possessions and have held the defense fixed. If we start attempting more attacks on the paint, we should expect second-order effects, such as the defense adjusting and bringing more help to defend the restricted area. A defender who’s been beaten off the dribble may be more willing to allow more space for a pull-up. A defender concerned with blowing his help-side defense for a second possession in a row may overcorrect and leave larger passing lanes and slower closeouts following a kickout.

This is the basketball application of Robert Lucas’s 1976 critique of economic policy evaluation. Historical relationships can change when a new policy changes how people behave. Here, the opposing team gets to react to the offensive “policy.”

It also explains why I would be careful with a chart showing teams lose accuracy when they increase rim frequency. It’s possible that pattern is consistent with diminishing returns. But things like roster changes, injuries, opponents, and changes in shot difficulty could also account for it. A team that increases both frequency and accuracy wouldn’t disprove diminishing returns either. That team may have improved its ability to create shots.

Got it. Don’t force every possession into the lane. Is that it?

Start with a small change and go to the film. Run an earlier PnR against a specific coverage. Attack the closeouts of a specific defender. Give a cutter more chances when his defender loses focus for a split second.

Then, compare the full-possession outcomes with the available alternatives, including kickout threes, shooting fouls, offensive boards, and turnovers prior to a paint touch ever happening. Separate half-court possessions from transition possessions, then match the half-court possessions by opponent, coverage, ball handler, and time left on the shot clock. For example, compare a Paolo paint attack with a mid-range pull-up against the same coverage. These matched comparisons still leave room for factors such as execution, fatigue, and injury to affect the result. Ultimately, we need to be able to explain why the possessions being compared would otherwise have had similar outcome expectations.

All this being said, the paint touch stat is still a good starting point. You always need a starting point. While it doesn’t give us the easy signal we may have initially expected (hey, 10 more paint touches, 2.6 more points!), it gives us the beginning of a framework for how to think about attacking the rim.

In a future post, I will look at a related problem: what can a league or cohort average tell us when the decision concerns one specific player or one specific decision.