Reading a sample

Reading a sample

A statistic on a match page is a count of things that already happened. Goals, points, and old meetings are records. They are not the next result written in advance. The next match is a new event, played by whoever is actually on the pitch.

A sample is the set of matches behind the number. Three games is a small sample. Thirty games is still one season, and it mixes easy fixtures with hard ones. A team can sit high in a table because of the opponents it has already faced. The average does not travel unchanged into the next opponent.

An average hides the mix. A side that scores three and then zero can show 1.5 goals a game, the same as a side that scores one and then two. The average is true and the matches were different. Look at the matches when the page gives them to you, not only the single figure.

A percentage needs a denominator. Four wins from five is not the same evidence as forty wins from fifty, even though both can be described as a high win share. Small denominators jump around. One result moves them a long way. Do not promote a short streak into a law.

Old meetings between the same clubs are a sample too, and often a tiny one. A two-one from three years ago was played by other players, in another season, sometimes at another ground. It is a fact about that day. It is weak evidence about this day.

Implied chance is a different kind of number. A decimal price of 2.00 is a shorthand for about a 50 percent chance before the bookmaker's margin. You can disagree with it. Your disagreement is only useful if you can say why this match differs from the price, in a way you could explain to someone else.

Keep your own settled results if you follow a list, and keep the prices. A long list of wins without prices does not tell you whether you were paid enough. Judge a sample that is long enough to include the dull weeks, and do not resize the story every time the last match flatters it.