Correlation versus causation in sports statistics

A hockey player takes more shots on goal in games where they play more minutes. The two numbers rise together. That's a relationship worth noticing, but it doesn't tell you that every extra minute caused another shot.

Correlation describes how measurements vary together. Causation is a claim about why a change happened. You can observe the first without having enough evidence for the second.

There may be a straightforward explanation, such as extra time creating more opportunities. There may also be other differences: power-play time, opponents, teammates or the game situation. The observed pattern doesn't separate those explanations for you.

In our invented example, a ten-minute appearance has two shots on goal. A thirty-minute appearance has six. The total triples, but both rates equal 0.20 shots per minute.

That helps explain one possible source of the larger total: there was more time to generate it. It doesn't show an improvement in the observed per-minute shooting rate, and it doesn't prove what would happen if the coach changed the player's minutes next game.

Two appearances are far too little to explain the whole relationship. The point of the calculation is to ask a better question. Did the player's opportunities change, their rate change, or both?

Suppose the player receives more minutes because their team is chasing a goal. That game situation can also lead to more attacking play and shot opportunities. Now an outside factor may be influencing both the minutes and the shots.

Or a move onto a power-play unit changes the type of minutes, not just their quantity. Treating every minute as identical would miss part of the explanation. A useful comparison looks at the role and situations alongside the totals.

These are possible explanations, not facts about our invented player. You would need the game records to investigate them. A statistical association alone doesn't confirm the story that sounds most convincing.

“The player recorded more shots in longer appearances” is an observed statement. “Giving the player twenty more minutes will add four shots” is a claim about a future intervention. The first statement doesn't establish the second.

When a report shows a striking relationship, ask what else changed and how the games were selected. A comparison of genuinely similar situations may help, while a carefully designed study can answer a different causal question.

You can still use a correlation as research context. The useful habit is to keep the observation and its explanation separate. That gives you a clear finding without pretending the cause has already been proved.