Start here · Lesson 6 of 10

Confidence and sample size

Every probability we publish carries a confidence interval — a range showing how much the number could still move as more evidence arrives. A wide interval means we are not yet sure, and we would rather show that than fake precision.

There is a subtlety we take seriously: forecasts made on nearby days cover almost the same stretch of the future, so they are not independent observations. We count that overlap and report an effective sample size that is smaller than the raw count. It is the honest denominator.

So when you see a number here, look at the interval and the effective sample size next to it. They tell you how much to lean on it.

See also: confidence intervalThe range around a probability showing how much it could still shift with more evidence — a wide interval means we are not yet sure., n_effEffective sample size — how many truly independent observations stand behind a number, after discounting overlapping periods that partly repeat each other., overlapping windowsForecasts made on nearby days cover almost the same future stretch, so they are not independent; we count that overlap so we never overstate the evidence.