An equation can fit a set of numbers and still fail to describe the physical phenomenon behind them. This article follows an ideal pendulum through two separate questions: whether a proposed expression has compatible units and whether its predictions agree with the evidence. An original example shows why dimensional rules can reject some formulas while admitting others that predict different periods. Research on symbolic regression then explains how learned proposals and explicit constraints can cooperate during the search itself. The discussion distinguishes recovering a known expression from discovering a new law, and reads recent performance claims with their actual denominators, noise conditions and supplied information. The central question is what remains unresolved after an equation has passed a check. Measurements, uncertainty and tests beyond the conditions used to construct a model remain essential to judging what the model can explain.
