How NBA Analytics Revolutionized Strategy but Created New Blind Spots

The piece uses the second-winningest coach in NBA history, Don Nelson — who died at 86 — to show what the metrics keep failing to capture: the connective geometry between players.

What the first-generation metrics actually fixed

Per the NYT analysis, points per possession normalized every offense to 100 possessions for true efficiency comparisons. True shooting percentage and effective field-goal percentage absorbed the free-throw line and the 3-point arc into a single signal — tying volume and efficiency to actual impact, where raw field-goal percentage had blurred them. Box plus-minus, assists and steals per 100 possessions — each metric built to answer one question: how good is this player, really?

That was the first correction. The midrange jumper was always inefficient; the 3-pointer had always been worth more. Box-score volume had always confused activity with actual impact. The math made it obvious.

The Goodhart collision, on court

The NYT piece frames the next problem sharply: once a model is built to evaluate individual output, behavior reorganizes around that model. The metric becomes the thing the player and the system optimize for. The measure stops being a measurement.

For a tactical reader, this is the structural filter that matters most right now. The number captures output. The film shows what was traded away to produce it — the passes that don't surface in assist columns, the spacing runs that collapse under the second defensive rotation, the off-ball movement that turns usage without generating a box-score line. A roster can score efficiently and still lose the connective geometry that makes the unit function. Every change affects everyone else; the metric can't see that.

The Nelson lens

Don Nelson, who died at 86, spent 11 seasons playing for the Boston Celtics and won five championships alongside Bill Russell. His coaching template worked because, in Nelson's framing, Russell was the second type of superstar — the one who makes the players around him look better than they are. Nelson built his systems to produce that same dynamic.

His structural moves predated any analytics framework: he invented the point-forward position, deployed three-guard lineups when the rest of the league still ran five traditional positions, and turned the 7-foot Nowitzki into a 3-point shooter covering guards two feet shorter. He built around what players made possible for each other, not what they produced individually.

His teams worked because the optimization target was team geometry. The current blind spot the NYT identifies is the gap between what possession-level models capture and what a roster needs to function as a system. Closing that gap means tracking the connective tissue on film, not just tallying the output on the box score.