Pattern coverage metrics
With 3 to 8 terms, the calculator targets three high-frequency families: arithmetic, geometric, and quadratic. Arithmetic detection checks constant first differences across all adjacent pairs. Geometric detection checks a stable ratio, avoiding division by values near zero. Quadratic detection verifies constant second differences, then solves a, b, and c.
Finite differences as evidence
A finite-differences table compresses a sequence into trend layers. Level 1 highlights additive growth; a constant row implies an arithmetic rule. Level 2 highlights curvature; a constant row implies a quadratic rule. This page prints up to three levels so you can validate the model visually, not just accept a label. For example, the sequence 1, 4, 9, 16 shows first differences 3, 5, 7 and second differences 2, 2. That constant 2 is twice the quadratic coefficient a. The chart overlays your fitted rule and reveals whether residuals drift upward or oscillate. Use it to spot breaks after a specific term or index very quickly.
Indexing and formula alignment
Many textbooks define the first term at n=1, while programming tasks often start at n=0. The Index Start control shifts the derived rule so that your first entered value aligns with the chosen n. This reduces off-by-one errors when you reuse the formula in homework, reports, or scripts.
Prediction window control
The Predict Next Terms input caps forecasts at 20 values to keep results readable and exports compact. For arithmetic, each step adds d. For geometric, each step multiplies by r. For quadratic, each step evaluates an·n² + bn + c at successive indices. The table lists both n and value for traceability.
Tolerance for real-world decimals
Measured data can introduce rounding, sampling noise, or unit conversions. The Tolerance setting relaxes equality checks using a relative threshold. If a sequence almost matches but fails, increase tolerance gradually, for example from 1e-9 to 1e-6. A larger tolerance may fit noise, but it can also mask genuine pattern changes.
Exports and reproducibility
CSV export captures the detected type, formula, parameters, input terms, predicted terms, and the generation timestamp. PDF export produces a clean, shareable summary suitable for submissions. Together they support reproducibility: a reviewer can rerun the same inputs and confirm the same rule and forecasts.