Five students report hours studied and test score: (1, 62), (2, 68), (3, 71), (4, 77), (5, 82). The dots climb in a nearly straight band. Lay a ruler across them so that it runs through the middle of the cloud, with dots on both sides, passing the point of averages (3,72). That ruler edge is a line of best fit. Through (1,62) and (5,82) it rises 20 points over 4 hours, a slope of 5, so y^=5x+57.
"Best" has a precise meaning. For each dot, measure the vertical gap between the dot and the line, square it, and add up the squares. The least-squares line is the one that makes this total as small as possible. Squaring stops gaps above and below from cancelling and punishes big misses heavily, which is also why one far-off point can tilt the line noticeably. The hat on y^ marks a predicted value, not a score any student actually got.
A straight line is a model, and it can be the wrong one. Hot cocoa cools fast and then slowly: 90, 66, 50, 39, 32 and 27 °C at 5-minute steps. Software will still fit a line, y^≈81.2−2.45x, but the dots sit above it at both ends and below it in the middle, the mark of a curve. The same line predicts about −17 °C after 40 minutes, colder than the room. Before using a trend line, check that the cloud is straight and that you are predicting inside the data's range.
Think of it like this. A line of best fit is like a straight road built through a scattered village: it cannot pass every house, so it takes the route that stays as close as it can to all of them at once. If the village curves around a lake, no straight road serves it well, and you need a curved model instead.
Core idea. A single straight line through a roughly linear cloud summarizes the trend: its slope is the typical change in y per unit of x, and its equation gives predictions that are trustworthy only within the data's range and only if the cloud is straight.