How many data points do you need for an accurate prediction?

A hidden trend exists. Each tick reveals one more data point. Watch the prediction improve and the uncertainty cone shrink as evidence accumulates.

Dots = known data. Line = prediction. Shaded cone = uncertainty. Dashed green = true trend.

What you're seeing

A chart with data points being revealed one at a time. A prediction line tries to forecast the future based on what's known so far. The shaded cone shows how uncertain that prediction is — wider means less confident.

How to read the graph

Orange dots are revealed data. The solid orange line is the current best-guess prediction. The shaded area is the uncertainty cone — everything inside it is plausible. The dashed green line (appears after a few points) shows the actual true trend. The red error chart at the bottom tracks how wrong the prediction is over time.

The key lesson

With 2-3 data points, you can predict almost anything — and you'll be wrong about almost everything. With 30+ points, the cone narrows dramatically. This is why early-stage conclusions are dangerous, why small studies mislead, and why 'I know someone who...' is the weakest form of evidence. More data doesn't just improve predictions — it reveals how wrong your early guesses were.

Data Revealed1 / 60

% Observed2%

Prediction Error94.1

Error Reduction0%

Cone Width±80

Horizon20 steps

Only 1-2 data points. The prediction is basically a guess — the cone of uncertainty covers almost everything. Like knowing one letter of a word.