Why Last Week's Number May Not Travel

A statistic is a description of what a particular group did under particular rules. Change the group and you do not get to keep the description for free. Sports conversations ignore this constantly. Last week's points per game is quoted after the leading scorer has been ruled out, as if the number were a property of the uniform rather than of the people who wore it.

The roster is part of the denominator

Points, goals, yards, and saves are produced by available players. If the player who created a large share of them is out, the team's recent rate described a roster that will not take the field. You can still mention the rate. You mention it with the absence attached, or you are answering a question about last week while pretending it is a question about tonight.

The same holds for a return. A team that struggled without its passer is a poor description of the team once the passer is back. The bad weeks are real history. They are not a current measurement of the restored group. Regression to the mean assumes a fairly stable process. A lineup change is a new process. Do not regress toward a mean that belonged to someone else.

Minutes and matchups travel poorly too

A player's points from a week of heavy minutes against a weak defense will not automatically appear in a week of reduced minutes against a better one. The number did not lie. It included conditions you have to restate. Home and away, rest, and opponent strength are the same kind of baggage. How to read home and away splits is one slice of it. Injuries are the slice that can zero out a contribution entirely.

Role changes matter even when nobody is hurt. A player moved from a starting role to a bench role has not become a worse athlete in the abstract. The opportunity changed. Quote the opportunity with the production, or expect to be confused next month.

Small news, large edits

One absence can be a small story in a deep roster and a large story in a thin one. The stat line will not tell you which. A team that spreads production across eight people loses less of its rate when one sits than a team that runs its offense through a single creator. Read the distribution, not only the team total. A total with no distribution is a black box, and black boxes travel worst of all.

Small samples compound the problem. Last week was already a short count. Remove a key player and the relevant count for tonight may be a handful of older games in which the same group actually played. Those games, not the shinier recent ones, are the closer comparison.

Keep the conclusion descriptive

Say what no longer applies, and stop. Last week's number described a different availability sheet. You do not need to replace it with a prediction, and you certainly do not need to replace it with a wager. This site's point is literacy. A legal sportsbook, where one is even allowed, is outside the job. Minors can learn the roster habit with a box score. They do not need a stake to make the habit useful.

Frequently asked questions

Can I still cite a team average after a star is injured?

Only with the injury in the same sentence. The average described the roster that produced it. Tonight's roster is a different measurement.

Do bad weeks without a player predict the weeks after the player returns?

They predict those weeks poorly. Use games in which the relevant group actually played, and remember that even those games are a sample.

Is lineup news a betting system?

No. It is a reason to stop treating an old number as current. It is not a price and not a recommendation to wager.