The Three-Pointer Has Become a Trap: The Hidden Price the NBA Pays for Its Analytics Revolution

In the 2026-25 season, a Western Conference team attempted 42 three-pointers per game on average — the highest mark in league history. It scored 118.3 points per 100 possessions, third-best in the league by that measure. Then it lost in the conference semifinals, in a series where its opponent attempted only 33 threes per game but recovered 29 percent of its own misses on the offensive glass. In my small studio in Shenzhen, I rewound the footage of that series four times. On the fourth pass, I stopped on a fourth-quarter possession in Game 6. The league's most three-happy team launched a corner three. The ball kicked out. Three players in blue jerseys were already back on the other end before the ball touched the rim. Three of them. One player stayed to rebound. The opponent grabbed the ball, ran, and seven seconds later hit a three going the other way — and made it. I called a friend who does analytics for an Eastern Conference team. He laughed. "You just noticed? We call it the three-point tax. Nobody sends an invoice, but every team pays it." That was the moment I realized the NBA's analytics revolution had created a debt nobody had entered on the balance sheet. To understand that tax, you have to go back to the beginning. In 2026, an NBA team attempted 18 threes per game on average. By 2026, that number was 35 — nearly double in fifteen years. The change didn't come from players shooting better. It came from a simple division every analytics department knows by heart: a made three is worth roughly 1.05 to 1.10 points on average, while a mid-range two is worth only 0.75 to 0.85. On the expected-value scale, the three wins. Always. Daryl Morey in Houston turned that arithmetic into an ideology. Drop the mid-range two entirely. Maximize threes and shots at the rim. Other teams saw the efficiency and copied it. Stephen Curry's Golden State proved a team could win a title on unprecedented three-point volume. By 2026, every team wanted to be a distorted version of the Warriors. Every gym taught the three. Every youth academy recruited kids who could shoot from deep. An entire generation of players grew up believing the three-pointer was the answer to every tactical question. And while the whole league chased the three, three things were forgotten. First, offensive rebounding. Long threes bounce far off the rim, and they bounce out toward defenders more than shots at the rim do. The NBA's offensive rebound rate fell from 32 percent in 2026 to 27 percent in 2026. A five-point gap compounded over 82 games is hundreds of lost possessions per season. Teams like Memphis or New York still kept a true glass-cleaning center, and they are exceptions worth studying. Second, transition defense. When five players shoot threes, five players are on the perimeter. When the ball caroms out, they are far from their own basket — and far from the opponent's basket in the worst possible way. Missed threes generate transition chances for the opponent faster than any other situation in basketball. Teams optimize for the shot, not for its consequence. Third, variance. Threes carry higher variance than shots near the rim. Over a long season, variance averages out into the equation. Over a seven-game series, variance averages out into an accident. And the series is where everything is decided. This is where public data fools even the smartest people. Look at a normal night. Team A attempts 40 threes, makes 14, a 35 percent rate. Team B attempts 20, makes 8, a 40 percent rate. Team B shot better by percentage. But Team A scored 42 points off threes; Team B scored only 24. Team A wins. The box score tells that story, and that story goes on television every night. But the box score does not count what happens after a miss. I took data from the last three seasons and split every possession into two parts: the shot, and the aftermath. On the shot, the three wins. On the aftermath, the three loses badly. Added together, the gap is not as large as the original arithmetic suggested — provided the team defends transition well and rebounds well. For a team that optimizes for the shot while ignoring the aftermath, the three becomes a net loss. The mistake of Moreyball was not the three-pointer. It was valuing the shot in isolation — as if a possession ended when the ball left the hand, rather than when the other team scored. I once fell into exactly that trap. In 2026, as a senior statistics student, I started a blog called "Hermes Perspective" to analyze CBA basketball data. I built a Poisson regression model for the road team's three-point shooting in the Southern Conference final between the Shenzhen Leopards and the Xinjiang Flying Tigers. The model predicted the road team should shoot eleven more threes per game. I wrote a piece titled "Why Must We Break the Bear's System?" My model was right about the shot. It was wrong about the series. That model is why I got an internship in Beijing, but it took me years to realize I had neglected half of the equation. The neglected half is where the game is actually decided. Look at Nikola Jokic's Denver for a different path. Jokic is not a high-volume three-point shooter. He shoots enough threes to be respected, not enough to live by them. Denver's attacking center is the high post, the pass into gaps a defender cannot ignore. Jokic scores through two-point efficiency, through passing, and through controlling pace. Denver does not optimize for one shot. It optimizes for the whole possession, start to finish, including the aftermath. Shai Gilgeous-Alexander's Oklahoma City is the second example. They shoot threes, but they also lead the league in generating steals and transition. Shai lives in the mid-range, which Moreyball ideology treats as worthless. But his mid-range is not isolated — it is framed by a top-two defense and a young, tenacious rebounding corps. People say the NBA is a league of stars. I don't believe it. The NBA is a league of dead-ball possessions — and nobody wants to admit it, because dead balls don't trend on Twitter. Try a different count. In an average 2026-25 game, there are about 100 possessions per team. Roughly 20 of them come from dead balls: offensive rebounds, steals, fouls, bad passes, inbound plays. Twenty percent of the game does not come from a set play. It comes from chaos. And teams that optimize for the eighty percent are wasting the twenty percent that decides playoff series. Here is the paradox: the more teams chase the model, the smaller the model's edge becomes. When everyone shoots 40 threes, shooting threes is no longer an edge. It is a minimum requirement. And once threes become a minimum requirement, the battle shifts to what few focus on: who rebounds better, who runs transition better, who stays composed in chaos better. Look back at the small-ball data. The Shenzhen Leopards' small-ball lineup in 2026 posted an offensive rating of 116.4 points per 100 possessions — 9.7 points higher than the starting lineup. That was what I found and wrote about. But I never looked at the other side: that same small lineup also allowed 119 points per 100 possessions. Its beautiful offense was eaten alive on defense. I dug through the data dumpster and found a diamond, then discovered the diamond had a flaw. Experience taught me that failed analysis is also worth publishing, sometimes more than successful analysis, because it shows a model's real limits. The basketball world likes success stories. I started to like the stories that show where the model is wrong. Getting a player's name wrong can still be fixed. Getting the tactics wrong costs you a loss. But getting the model wrong — valuing the shot correctly and the possession incorrectly — costs you an entire season, and nobody enters it on the balance sheet. There is one detail I have not forgotten. In June 2026, when I went to Moscow as an on-site commentator for the Mexico-Germany World Cup match, I mispronounced Hirving Lozano's name as "Lozanho" three times and was corrected on air. After the match, I sat down and watched all 42 Mexico possessions on tape, and discovered that a 4-4-2 with pinched fullbacks had shattered Germany's defense. I wrote "My Mistake, and Löw's Mistake," using an xG model to show Mexico generated more dangerous shots through a high press. That day taught me two things: admitting a mistake publicly does not kill your credibility, and data only has value when placed in full context — not one shot, but a whole move; not one three, but a whole possession. The NBA is at exactly that inflection point. The wave of copying Moreyball has reached saturation. Every team knows what a three is worth. The marginal edge now lies in what simple models cannot count: the quality of the possession after a miss, the ability to control pace, and the rebounding instinct an entire generation of players was taught to forget. Next season, I will track a metric no broadcaster displays: the points a team allows the opponent to score within seven seconds of its own missed three. I call it the "recovery tax." It is not in the box score. It does not appear in highlights. But I believe it predicts a series better than three-point percentage. Heresy today, orthodoxy tomorrow — I only bet one step ahead of everyone else. And this step, I am betting on the teams willing to pay less recovery tax. An empty arena does not kill basketball; it only strips the makeup off the sophists, and the empty arena in my head is a data table nobody has bothered to read to the end. When an NBA team accepts giving up five percent of its three-point volume to win back five percent of its offensive rebounding, that team will win a championship. The question is not whether anyone dares. The question is who does it first — and whether I can bet before the whole league realizes it has been paying that tax for fifteen years.