Simple baselines that are harder to beat than you'd expect
A moving average forecast simply predicts the next value as the average of the last N observed values, sliding that window forward as new data arrives. It's crude, it lags behind sudden changes in trend, and it treats every observation in the window as equally important, but it's also fast, interpretable, and often a surprisingly tough baseline to beat on noisy series with no strong structure. Exponential smoothing improves on this by weighting recent observations more heavily than older ones, with the weight decaying exponentially the further back you go, so it reacts faster to genuine changes while still smoothing out noise. Extended versions (Holt's method, Holt-Winters) add explicit components for trend and seasonality, letting a fundamentally simple technique handle fairly rich patterns.
The appeal of these methods isn't that they're the most accurate option available, it's that they're cheap to compute, easy to explain to a stakeholder, and remarkably robust. Any forecasting project should establish one of these as a baseline before reaching for anything more sophisticated, because a complex model that can't beat a moving average isn't earning its complexity.