How this calculation works
Changing the spacing between observations changes the interpretation. The rule assumes the steps are equally spaced.
Next term = 3u₃ − 3u₂ + u₁A worked example
For 2, 5, 10, first differences are 3 and 5. The next difference is 7, giving 17.
Good to know
Three observations alone do not establish a unique general sequence.
Reference: OpenStax Algebra and Trigonometry (opens in a new tab) ↗
Extend the second-difference row by one position
The observations 2, 5 and 10 have first differences 3 and 5. Their second difference is 2. Assuming that second difference remains constant, the next first difference is 5 + 2 = 7, so the next sequence term is 10 + 7 = 17.
Algebraically, the new first difference is (u₃ − u₂) + [(u₃ − u₂) − (u₂ − u₁)]. Adding it to u₃ gives the shortcut 3u₃ − 3u₂ + u₁. The triangular SVG aligns each difference underneath the pair of terms that produces it.
State the polynomial assumption behind the prediction
The rule assumes equally spaced observations and a polynomial of degree at most two. A linear sequence has second difference zero and is included as a special case. For unequal x intervals, the Quadratic through three points calculator uses divided differences instead.
Three terms cannot prove how a sequence continues. At indices n = 1, 2 and 3, the polynomial n² + 1 gives the observations 2, 5 and 10. Adding K(n − 1)(n − 2)(n − 3) leaves those three values unchanged but changes the fourth to 17 + 6K. Any desired fourth value can therefore fit the same observations under a different rule. Use the displayed answer as a model-dependent extrapolation and compare it with further observations before trusting a trend.
A few questions, answered
Can the predicted term be negative?
Yes. The inputs and their finite differences can have either sign.
How can I check the constant-second-difference assumption?
Obtain additional equally spaced observations, form their difference rows, and compare the observed second differences.