Complex Interpolation and Extrapolation (IOD 603)

A straight-line trend grows by the same amount every step, so stretching it forward adds one more equal-sized chunk each time. A curved trend changes by a growing or shrinking amount each step, so when you extend it, you have to keep matching that changing rate, not just add a flat amount. Miss the curve and assume a straight line instead, and the gap between your guess and the real curve grows fast, because the curve's own rate of change is accelerating or decelerating while your flat guess sits still.

To extend a bending curve, look at how the curve is bending right at its last point, and continue that same pattern of change. If a count doubles every step, keep doubling for each step beyond the data. If a quantity grows with the square of another, keep squaring, not multiplying evenly. Read the pattern from at least three or four points before trusting it forward even one step, and treat every extra step past the last measurement as another multiplication by that same pattern, not an extra flat jump.

This is where extrapolation gets risky. A curve that fits four data points beautifully can still be the wrong shape once you leave the data, and the farther out you project, the more that hidden error compounds. Worse, real systems often have a physical ceiling, limited food, limited space, limited fuel, that no equation knows about. The trend line only describes what happened inside the measured range; past that range, matching the curve's shape is necessary but not sufficient, because the real process can hit a wall the numbers never warned you about.