Seventy Points in a Loss: The Biggest Gap in Basketball Data

The clock at TD Garden read 7:30 p.m. on March 24, 2026, when the ball went up. In Hanoi it was 6:30 the next morning, and I was already in front of a screen with a paper notebook — a habit left over from the days when I watched V.League tapes with editors from an older generation, back when every judgement had to start with a hand count.

Forty-five minutes later, Devin Booker left the floor with 70 points. The Phoenix Suns lost 120-130 to the Boston Celtics.

The individual box score from that night reads like a monument: 21-of-40 from the field, 24-of-26 from the line, eight rebounds, six assists. He was 20 years old, in his second season. The team's result column was blank. For nearly a decade since, I have used that night as the entry test for anyone who wants to learn how to read basketball data: what do you see first, the stat line or the result?

Vietnamese fans now reach a box score faster than ever. A phone app returns the numbers within minutes of the final horn, and it drags behind it a flood of commentary written before anyone has rewatched a single possession. That speed helps you follow the league, but it also builds a damaging habit: reading a stat line the way you read a verdict.

In my trade, the label “empty stats” was born out of exactly that gap. It does not describe a wrong number. It describes a correct stat line produced under conditions where the value of winning is zero — the team is long eliminated, the playoff race is over, and most of the minutes come after the outcome is settled.

Whenever I profile a player, I run through nine dimensions of data: tactics and technique, individual profile, salary structure, league landscape, rules and governance, coaching and locker room, risk, media narrative, and industry ripple. Those nine dimensions can reconstruct a season. But the biggest lesson I have taken from that process sits somewhere else entirely: what happens when one of the nine has no input data at all.

I learned that answer through a failure. In 2026 I built a group-stage model for the World Cup using expected goals, actual goals and possession control, then concluded Germany would advance because they carried the highest accumulated attacking output in their group. Germany went out in the group stage. Looking back, I found what I had never collected: Japan posted a PPDA of 6.8 across their matches against Germany and Spain — pressing intensity in the most suffocating bracket of the tournament. Their pressing lived entirely outside my data set, so the model had no way to see it. It took me three months to build a system that integrates non-traditional sources, and since then every analysis carries its own section: risks and gaps.

Back to that night at TD Garden. The first thing I did when I reopened the tape was compute the true shooting efficiency of that 70-point line. The familiar formula: points divided by twice the sum of field-goal attempts and 0.44 times free-throw attempts. With 40 attempts and 26 free throws, the denominator is 102.88. The result: 68.0% TS. The NBA league average that season sat around 55%.

Put differently, that stat line was never empty in efficiency terms. It was far above the league standard. Booker was not chucking to reach a milestone; he scored efficiently at an enormous workload, and most of the points came out of structured possessions.

So why does the “empty” label still stick? Because that label does not measure efficiency. It measures context. Phoenix finished the 2026-17 season at 24-58, last in the Western Conference. When a team in that position reaches March, every game has stopped meaning anything in the standings. What remains on the floor is development time for young players, and a coaching staff being judged by a different question altogether: has this franchise found someone to build around?

This is where my framework splits from the conventional read. I do not ask what the stat line looks like. I ask four questions, and all four need concrete inputs.

The first is usage rate. For Booker that night it was at its ceiling — he was the destination of nearly every second-half possession. High usage is not inherently bad; it only means the value of the stat line depends entirely on efficiency, and here efficiency had already answered.

The second is the on/off differential. If Phoenix were dramatically worse whenever Booker sat, the line carries structural value. If the team performed about the same without him, the story flips. This is a metric Vietnamese media almost never cites, partly because it is hard to look up, and partly because it destroys tidy narratives.

The third is the garbage-time split. This is the most abused filter in the business. In principle, garbage time is the stretch after the result is settled and both teams have emptied their benches. In Booker's case, most of the scoring came while the game was still being watched, with Boston's starters still on the floor. The garbage-time filter does not clean that line.

The fourth, and the most ignored, is the team's incentive structure. A 24-58 team does not behave like a team chasing a playoff berth. A coaching staff letting a young player take 40 shots is not doing it because they believe it produces wins, but because they need to find out who he is. That is an investment, not a show.

Stack the four answers together and the picture is far clearer than the label. Efficiency was elite, usage was maximal, the garbage-time filter does not apply, and the team's incentive was development. Three of the four dimensions support taking the line seriously. The remaining one — the final score and the standings — is what pushes it into the “empty” folder.

Which leads to the conclusion I consider the single most important point in this whole subject: “empty stats” is a judgement about context, not a judgement about quality. People use it as though the two were the same thing. They are not.

I learned to separate them on another night, three months earlier, on a football pitch a few hundred kilometres from Hanoi. In 2026, while working as a data editor for a football site, I wrote that Hanoi FC deserved to win 3-1 rather than scrape a lucky 1-0 against Quang Nam. I based it on expected goals of 2.87 against 0.45, 68% possession, and 14 shots inside the box. The piece was mocked, and a week later coach Chu Dinh Nghiem admitted he had rewatched the tape and adjusted his tactics based on that analysis. That night, the media called them soulless. xG said the opposite, and I chose to trust xG.

That trust was not free. In 2026 I left Hanoi for Russia, and while most of my colleagues picked Brazil or Germany, I wrote that Croatia would reach the final. The basis was not inspiration. Their midfield covered an average of 112 kilometres per match, the highest at the tournament, while the trio of Modric, Rakitic and Brozovic held a PPDA of 8.2. Croatia did not reach the final because of luck. They reached the final because of legs that would not stop. The international reporters who laughed at that prediction were the same ones who later introduced me to an Opta analyst.

Bring that principle back to basketball and the same error repeats: people judge a stat line by the final score instead of by the structure that produced it.

But stopping there would have built me a different trap, and I have fallen into it often enough to know where it sits. That trap is the mirror image of the earlier mistake: calling every development rep an empty stat simply because the team is losing.

A 20-year-old on a last-place team has no other laboratory than the game itself. If the coaching staff does not let him take 40 shots during a lost season, they walk into the next season without knowing what they own. Slapping the “empty” label on nights like that quietly becomes a refusal to invest in development — and in basketball, where the learning curve for young players is steeper than in any other team sport, that refusal has a price.

In the other direction, there is a category of cases where I genuinely distrust the stat line. That is when a team has a clear market incentive: a contract about to expire, an All-Star campaign being lobbied for, or a trade package waiting to be wrapped. Then high usage stops being development and becomes a product manufactured for sale. Telling those two situations apart requires contract structure and cap data — exactly the dimension Vietnamese readers find hardest to reach.

Then there is my own mistake. When the stands went empty, my model collapsed. I knew I had forgotten the human factor. In 2026, as the Bundesliga returned to empty stadiums, the home-advantage model I had built since 2026 predicted the home win rate would fall from 54% to below 50%. The actual figure was 48.7%, and Borussia Dortmund won only three of their remaining eight home games. The prediction was right, but the recovery model that followed failed badly, because I had not accounted for differences in training-ground quality and squad psychology. Data shows a trend, not a prophecy.

That is why I no longer use the phrase “decisive metric”. Basketball has regions data can touch but cannot hold: the noise in a locker room, an ankle that has not fully healed, a coach losing his dressing room. A good model is one that can name its own blind spots.

Four years after that night at TD Garden, Phoenix reached the 2026 NBA Finals with Chris Paul running the offence and Deandre Ayton anchoring the paint. Booker scored 42 points in Game 4 and 40 in Game 5, the latter a Phoenix win that cut the series deficit against the Milwaukee Bucks. They lost the series 2-4. The “empty” label disappeared, but it disappeared because the context changed, not because the old stat line was rewritten. The stat line had been right all along.

What I want you to carry away is not a conclusion about Booker. It is a way of asking questions. The next time a player scores 40 in a loss, check true shooting efficiency first, then usage rate, then ask what that team is actually trying to do with the rest of its season. If the three answers agree, the stat line is telling a real story. If they contradict each other, you are standing in front of a data gap — and that gap is worth writing about rather than papering over with a label.

I do not believe in hunches. But I believe in what hunches confirm once the data validates them. Numbers never need us to defend them. We need them so we stop lying to ourselves.