Expecting the Next Shot: What Basketball Analytics Really Knows

A run of threes

In Game 7 of the 2018 Western Conference Finals, the Houston Rockets kept launching three‑pointers.

For months that approach had been one of the most efficient offenses in basketball. Houston had built an entire system around spacing the floor, driving into the paint, drawing fouls, and firing threes whenever defenders hesitated. Over a long season the math worked. Possession after possession, the returns accumulated.

But in that decisive game the shots stopped falling. One miss turned into two, then five, then ten. By the time the streak ended, Houston had missed twenty‑seven consecutive threes, and the season slipped away.

For many observers the moment felt like a verdict. Maybe the numbers were wrong after all. Maybe the whole analytics project had misunderstood the game.

Yet the strange thing about that night is that it reveals something deeper about analytics rather than refuting it. To see why, it helps to borrow a question from the eighteenth‑century philosopher David Hume: when we expect the future to resemble the past, what exactly justifies that expectation?

What statistics actually promise

Hume noticed something unsettling about everyday reasoning. We constantly move from what has happened before to what we think will happen next. The sun has risen every morning we remember, so we assume it will rise tomorrow. A shooter has hit thousands of jumpers in practice, so we expect the next one to look familiar.

But if you examine the logic carefully, the step from past to future never becomes a proof. No list of past examples, no matter how long, can logically guarantee that the pattern must continue. This is what philosophers call the problem of inductionThe problem of induction, first articulated by David Hume, is the challenge of justifying our expectation that the future will resemble the past. No amount of past evidence can logically prove that a pattern must continue. . The future might still surprise us.

That does not make the expectation irrational. It only means that the expectation is probabilistic rather than certain. Repeated experience builds confidence, but it never transforms confidence into necessity.

Basketball analytics lives exactly in that territory. When analysts describe efficient shot profiles or lineup advantages, they are not discovering mathematical laws that dictate what will happen in the next game. They are organizing experience — thousands of possessions, seasons of shot data, patterns across teams — into stronger and weaker expectations.

In other words, analytics is not about certainty. It is about disciplined expectation.

When patterns reshape the game

Consider how the modern three‑point revolution unfolded.

The 2004–05 Phoenix Suns played at a blistering pace, spread the floor with shooters, and produced an offense that led the league in scoring. At the time the style felt unconventional, almost experimental. Yet over the course of the season the same pattern appeared again and again: space the floor, attack quickly, and the defense eventually cracks.

Repeated success began to change belief.

Years later the Houston Rockets pushed the logic further, attempting over forty threes per game during their sixty‑five‑win season. The strategy was not based on the idea that every three‑pointer was good or that every drive would succeed. Instead it rested on a more modest claim: over large numbers of possessions, certain shot types tend to yield better results than others.

That modest claim is exactly the kind of inference Hume described. Analysts observe the same conjunction repeatedly — spacing leads to open threes, open threes produce efficient offense — and expectation gradually forms. CustomIn Hume’s philosophy, custom (or habit) is the psychological mechanism by which repeated experience produces expectation. We come to expect patterns not through logical proof but through the force of accumulated experience. , built from data rather than intuition alone, begins to guide strategy.

Eventually the pattern becomes league knowledge. By the 2023–24 season, the Boston Celtics led the NBA in both three‑point attempts and makes while producing the league’s most efficient offense on their way to a championship. What once looked radical had become routine.

Yet even this success does not create a universal law. It only strengthens the expectation that spacing and shooting usually work well.

Usually is the key word.

Good reasoning, bad nights

This brings us back to Houston’s Game 7 collapse.

When viewers watch twenty‑seven straight misses, it is tempting to treat the sequence as proof that the strategy itself was flawed. But from a probabilistic perspective the night demonstrates something different: even well‑supported expectations leave room for variance.

A player can take a high‑quality shot and still miss. A team can construct an efficient offense and still lose a single game. The fact that a probabilistic strategy fails in one instance does not mean the reasoning behind it was mistaken.

Hume would say that the Rockets’ system rested on strong inductive evidenceInductive reasoning draws general conclusions from specific observations. Unlike deductive proof, inductive conclusions are always probabilistic — they grow stronger with more evidence but never reach absolute certainty. — months of data, thousands of possessions, repeated success across opponents. What the Game 7 streak exposed was not bad logic but the unavoidable gap between likelihood and certainty.

Analytics often gets criticized for moments like this because people quietly treat statistics as if they promise guarantees. When the guarantee fails, the numbers seem discredited.

But the numbers never promised that kind of certainty in the first place.

Forecasts, not prophecies

Think of a weather forecast that predicts an eighty‑percent chance of rain. The forecast is useful precisely because it acknowledges uncertainty. If the sky stays clear, we do not conclude that meteorology has collapsed. We simply recognize that probability always allows for exceptions.

Basketball analytics works the same way. A spacing offense may create the most efficient shot distribution in the league, yet a cold shooting night can still derail it. A dominant shooter like Stephen Curry can build a season of historically accurate three‑point shooting and still miss several in a row when the pressure rises.

Those moments feel dramatic because they interrupt expectation. Yet expectation itself — the belief that Curry will probably make the next one, or that an open three is generally a good outcome — still rests on the accumulated evidence of countless earlier possessions. The reasoning is probabilisticProbabilistic reasoning assigns degrees of likelihood to outcomes rather than treating them as certain or impossible. It acknowledges that even well-supported expectations can fail in individual cases while remaining reliable over large samples. rather than certain.

Hume’s insight is that such expectations are unavoidable. Coaches, players, and fans all rely on them. Scouting reports, lineup decisions, and even casual predictions about the next shot all depend on the same basic leap from past experience to future possibility.

Analytics simply performs that leap with wider evidence and greater care.

Seeing the numbers differently

Once you view analytics through this lens, many familiar debates look different.

Arguments about “watching the game” versus “trusting the numbers” often assume that one side deals in experience while the other deals in abstraction. In reality both sides are interpreting experience. The difference lies in how broadly the evidence is gathered and how cautiously the conclusions are framed.

The most thoughtful analytics writing rarely claims that a strategy must succeed. Instead it asks what tends to succeed over time, what evidence supports that expectation, and how confident we should be in the pattern we see.

Hume’s philosophy helps explain why that modest approach is not a weakness but a strength. The future of a basketball game can never be proven in advance. At best we form expectations shaped by past patterns and revise them when new evidence appears.

The three‑pointer that rims out in a decisive moment does not invalidate the reasoning that produced it. It only reminds us that probability governs the sport more than certainty does.

And perhaps that is part of why basketball remains compelling: the patterns are strong enough to guide belief, yet fragile enough that the next possession can still surprise us.

---

Footnotes / Philosophy Terms

1. Problem of induction ↩

The problem of induction, first articulated by David Hume, is the challenge of justifying our expectation that the future will resemble the past. No amount of past evidence can logically prove that a pattern must continue.

2. Custom ↩

In Hume’s philosophy, custom (or habit) is the psychological mechanism by which repeated experience produces expectation. We come to expect patterns not through logical proof but through the force of accumulated experience.

3. Inductive evidence ↩

Inductive reasoning draws general conclusions from specific observations. Unlike deductive proof, inductive conclusions are always probabilistic — they grow stronger with more evidence but never reach absolute certainty.

4. Probabilistic ↩

Probabilistic reasoning assigns degrees of likelihood to outcomes rather than treating them as certain or impossible. It acknowledges that even well-supported expectations can fail in individual cases while remaining reliable over large samples.