This story is part of Peak, The Athletic’s desk covering the mental side of sports. Sign up for Peak’s newsletter here.
Joe Boylan worked in the NBA for a decade with the Minnesota Timberwolves, New Orleans Pelicans, Memphis Grizzlies, Golden State Warriors and Boston Celtics. He is the co-founder of Cognition Coach.
There is a principle in economics called Goodhart’s Law: When a measure becomes a target, it ceases to be a good measure. It was formulated to describe monetary policy. It describes basketball better than almost anything else.
It also describes almost every organization that has ever tried to manage what it measures.
Don Nelson spent 11 seasons playing for the Boston Celtics. He won five championships. He watched Bill Russell up close every day for more than a decade.
When Russell died in 2022, Nelson said, “There are two types of superstars. One makes himself look good at the expense of the other guys on the floor. But there’s another type who makes the players around him look better than they are, and that’s the type Russell was.”
Nelson would go on to become one of the winningest coaches in NBA history, in large part by building teams that purposely challenged the established perspective. He invented the point-forward position. He ran three-guard lineups when everyone else played traditional big men and stuck to the five traditional positions. He convinced 7-foot Dirk Nowitzki to shoot 3-pointers and guard players 2 feet shorter.
When you change the way you look at something, the things you see change.
Nelson’s teams worked because he built around what players made possible for each other, not what they produced individually.
A basketball team is a complex system. Every change affects everyone else. The analytics revolution would eventually give the sport a language to describe many of Nelson’s instincts. But it also created a new problem, one that sits at the center of how good teams function. It’s not just a basketball problem.
The analytics revolution would eventually give the sport a language to describe many of Nelson’s instincts. But it also created a new problem, one that’s not just a basketball problem. The moment you pick a number to measure success, behavior quietly reorganizes around that number.
Every organization that has ever tried to measure performance faces the same blind spot. Basketball is just an unusually honest place to find it.
The analytics revolution in basketball is one of the most important intellectual shifts in sports history. It corrected decades of bad intuition inherited from an era predating the 3-point line. It helped teams see that 3-pointers counted for more than midrange jumpers, and that box-score stats were hiding more than they revealed.
Some of the changes were software updates that make obvious sense. Points per possession replaced points per game, normalizing every offense to 100 possessions to actually compare efficiency. True shooting percentage and effective field-goal percentage are clearer versions of raw field-goal percentage because volume and efficiency from the free-throw line and the 3-point line are more directly tied to actual impact. Box plus-minus. Assists and steals per 100 possessions. These first-generation metrics were built to answer a specific question: How good is this player, really?
That work was real, and it mattered. These were the right questions for that era.
Then Goodhart’s Law kicked in. Once teams built models to evaluate individual output better, everything was organized around optimizing for that target. Players narrowed their offensive roles, giving the best players more control. A high-usage pick-and-roll handler. A lob threat and offensive rebounder. The idea of a 3-and-D player. Offenses became cleaner, more optimized to stretch the defense: run to the corners, make the reads clear.
These were meaningful steps forward, but Goodhart’s Law is an argument for watching what happens to behavior once you start. And what happened in basketball is that optimizing for individual metrics began to obscure a different kind of value, the kind Nelson had been coaching toward for decades.
The dynasty of the Golden State Warriors is the clearest case of what individual metrics fail to capture.
Klay Thompson became one of the greatest shooters of all time, next to Steph Curry. But what he didn’t do was arguably just as important. His inability to dominate the ball meant he never did. He shot when open, passed when not. The ball never stuck in his hands, forcing his defender to guard him in unfamiliar positions and allowing other players to benefit from his gravity. Thompson’s limitation became a team strength that no individual efficiency metric was designed to capture.
His teammate on Golden State, Draymond Green, had the opposite limitation: He couldn’t really shoot. That made him think pass first. He became the best playmaking forward in basketball precisely because scoring wasn’t an option. The defense couldn’t ignore him in the pick-and-roll because if they did, he’d find the open man.
Playing alongside Curry made Green more valuable, and Green’s passing made the system around Curry more valuable. That circularity is the whole point of basketball, and it’s exactly what makes this so hard to measure.
First-generation models, built to evaluate individual offensive creation, undervalued both Thompson and Green. What those metrics couldn’t see was how their limitations created conditions for everyone to thrive. This is Nelson’s principle at scale: The value isn’t in what each player produces. It’s in what they make possible for everyone else.
Every organization sees this in its indispensable personality hire.
The 2014 San Antonio Spurs told the same story from a different angle.
Gregg Popovich’s championship team is remembered for its ball movement, which people often call “the beautiful game.” Tim Duncan, Tony Parker, Manu Ginobili and a rising Kawhi Leonard shared the main load. However, the roster around them was not filled with prototypical players every team chases.
The championship Spurs didn’t get to the free-throw line (they were 27th in attempts). They didn’t generate extra possessions (they were 24th in offensive rebounding). They ranked first in offensive efficiency anyway.
Part of that was talent. Three future Hall of Famers and a budding superstar make a lot of things work. However, something specific was happening with the way the pieces fit together, particularly off the bench.
The Spurs’ second unit didn’t have much off-the-dribble creation outside of Ginobili. Marco Belinelli, Patty Mills and Matt Bonner were all viewed as defensive liabilities. Boris Diaw was a low-volume 3-point shooting playmaking big with an outsized AST% — a pace-adjusted metric for playmaking impact rather than raw assist numbers.
Individually, none of these players jumped off the page. Together, they were nearly impossible to stop.
The tools to measure some of this already exist. Screen assists are now tracked, so the screener finally gets credit. Gravity scores measure how much defensive attention a player commands without the ball: How much space he creates that never shows up in a box score. Regularized adjusted plus-minus (a model that estimates a player’s impact on team performance while controlling for who else is on the floor) has gotten increasingly sophisticated. The most promising work is in lineup-adjusted individual value: The idea that a player’s worth changes depending on the system around him.
Yet most teams still don’t build around these.
This is where Goodhart’s Law hits hardest. The problem isn’t that we can’t see collective value. It’s that the entire infrastructure of basketball decision-making (contracts, trades, draft boards, media narratives) is still organized around individual output. A player who passes up a good shot to create a great shot for a teammate looks less efficient by the metrics that determine his salary. A player whose gravity creates space but never touches the ball makes everyone else more efficient. His stats look empty compared to the always-around-the-rim big man gobbling up rebounds and blocks. His contract looks like an overpay. His value is enormous.
Nelson understood this decades before anyone had a spreadsheet to prove it.
He watched Russell make everyone around him better and stored it away. Nelson didn’t have the data to prove it. He didn’t need it. He built teams around it — point-forwards, three-guard lineups, a 7-footer launching threes — because he understood that a player’s value lived in what he made possible, not what he produced.
The first analytics revolution gave basketball the language to challenge inherited thinking about individual players. The next one has to measure the whole sentence.