Reading a series

Time series data is just values in order: daily prices, hourly temps, monthly sales. Order matters. The usual split is trend (long drift), seasonality (repeats on a fixed cycle), and noise (the leftover).

Trend

A linear fit gives slope (change per step) and R² (how much variance that line explains). Strong slope + decent R² usually means a real drift, not a fluke — still check a plot before you trust it.

Seasonality and smoothing

Seasonality is the repeating bump (period 12 on monthly data often means annual). Exponential smoothing pulls a quieter line through the points; a higher alpha trusts recent values more, a lower alpha stays smoother.

Practical tips

• Aim for 10–15+ points for a crude linear trend; more is better
• Seasonality needs at least a couple of full cycles
• Investigate outliers before deleting them — error vs real spike