Houston, the 27th Miss and the Empty Column

On the 27th miss of the streak, the ball sat in Trevor Ariza's hands in the left corner of Toyota Center. Nobody in the front row stood up. Ariza gathered the ball, held his eyes on the rim one beat longer than usual, and shot. The ball hit the front iron and kicked out. I was sitting in a studio in Chicago, my headset still carrying the shrill voice of a colleague calling James Harden's name, but my eyes were fixed on the Houston bench. Mike D'Antoni stood there, hands in his pockets, and did not call timeout. Nobody got angry. Nobody clapped. That team was doing exactly what its report demanded, and that is precisely what sent a chill down my spine.

On May 28, 2026, the Houston Rockets entered Game 7 of the Western Conference Finals with 65 regular-season wins, the best record in the league and a franchise mark. They led 54-43 at halftime. They lost 92-101. For the game they shot 44 threes and made 7, including a run of 27 consecutive missed three-point attempts, the longest ever recorded in a playoff game. Chris Paul sat out with a hamstring strain sustained in the final seconds of Game 5, when the result was already settled.

That season Houston took 3,470 three-point attempts, most in the league, and that was the foundation of an entire philosophy built by Daryl Morey: basketball holds only two shots worth taking, at the rim and beyond the arc. Everything in between was treated as waste. The revolution lifted Houston from a mid-table team to a title contender inside four seasons, and it was so persuasive that other teams copied every detail, including the ones they did not understand.

After that night, American basketball split in two. Half said Houston lost because of Paul's hamstring. The other half said Houston lost because of 27 missed threes, and that if only four of them had dropped, the story would be different. Both explanations are tidy, both are comfortable, and both skip the thing I care about most: inside that organisation, at halftime, nobody had the authority to say the words “I don't know”.

Data engineering has a concept called the null gate. When an input field returns an empty value, the system must stop and report “cannot assess”, instead of taking a default and running on. It sounds trivial. Yet most disasters in sports analytics are disasters of the missing gate.

Reconstruct Houston's tracking board using the nine standard categories any analytics department keeps, and by the second half eight of them returned empty. The second rhythm creator behind Harden, empty, because Paul was in a t-shirt at the end of the bench. A mid-range option against a defence switching everything, empty, because Houston had not practised it all season. Offensive rebounding, empty. A plan for a half in which the legs are gone, empty. And the system did exactly what a system without a null gate does: it took the default. Houston's default was to shoot more threes.

Houston did not lose its composure while missing 27 straight threes. It took 27 straight threes because its system had no room for an empty cell.

In that second half, Houston scored 38 points and Golden State scored 58. D'Antoni's rotation shrank to seven men, and three of those seven played more than 40 minutes. Ariza left the floor with 0 points on 12 shots. Harden scored 32 but had to manufacture almost everything himself. Golden State had switched one through five for two seasons, Houston knew it in advance, and still had no second option, simply because a second option was not worth anything inside the model. Every time Harden caught the ball, the Warriors' defence exhaled one beat, a beat anyone sitting close enough to the floor could hear, and no column on the board recorded it. I have followed Morey since the years he used spreadsheets to argue with scouts, and I have always admired the way he turned a belief into a machine. The better the machine, the harder it becomes to hear someone say the model is missing a variable.

That is why I do not trust conclusions that arrive pre-packaged. A good analyst can tell me Ariza's corner three was worth 1.08 expected points, and he is right. But he cannot tell me what that same shot was worth at minute 40 of a Game 7, after 26 misses. The model is not wrong in how it calculates; it is wrong in never returning “cannot assess” once the humans are spent. A 60-million player is not guaranteed to make more difference than a shy kid from an academy who knows how to watch, and that night, the only man who might have seen the empty cell was sitting on the bench in a suit.

I have seen that missing gate somewhere else. In 2026 in Kazan, the German national team left the World Cup with a statistical sheet better than its opponents in almost every column, and nobody on the coaching staff dared say those columns were measuring the wrong thing. Germany had probably lost before the first ball was kicked, people simply did not have sharp enough eyes to see it. Houston did not lose that way. Houston lost the way an organisation loses when it is so good that there is no room left for uncertainty.

I have to claim my share too. People like me read stat sheets on air as though they were truth, because they are compact, because they fit four seconds of broadcast, because nobody wants to hear a commentator say “I don't have enough data”. We teach audiences to trust the column, and then when the column is empty, we fill it with our own voice.

There is a stronger reading than mine, and I have to state it. Houston lost because Houston missed. Going 0 for 27 is a rare event, but rare events still happen, and if Ariza had made two of them this article would not exist. Morey's process produced 65 wins and an 11-point halftime lead against the best team of a generation; calling that process broken is too convenient for the man writing. The cause may sit somewhere else entirely: Paul went down in the last minutes of Game 5, and no model replaces the loss of a playmaker at that level. If someone tells me Houston lost purely because of a hamstring, I have no data to argue with.

But one thing I will hold. The difference between “the model was right and the result was bad” and “the model had no capacity to declare itself short of data” is the difference between an accident and an accident with no brakes. Every team eventually loses a game like this one. Very few teams have someone permitted to stand up at halftime and say: eight of our nine categories are empty, we have to play the second half some other way. Every giant's failure is a slap for those who collect names instead of collecting people.

What I want to see in the next three seasons: an NBA team hiring a dedicated director of missing data, a person whose only job is to say out loud that the cell is empty. My testable prediction: before the 2028-2029 season, at least one team will publicly release an uncertainty report ahead of a playoff series, and at least one coach will lose his job for daring to read it. For three years we chased a ball that seemed to have no defender, and it turns out what we were chasing was the silence in the middle of people. And if next season you watch a team take 40 threes and lose, ask them one question before you ask about percentages: at halftime, did anyone dare say that column was empty?