Addition chain generation underlies fast exponentiation in public-key cryptography. Decades of research have produced a family of classical constructive heuristics -- deterministic, polynomial-time methods that build a chain directly rather than searching a solution space. This paper reviews the major classical heuristics (binary/square-and-multiply, m-ary/window, sliding-window, factor method, power-tree, vectorial/Euclidean addition chains, and addition-subtraction chains), formally specifying for each its input parameters, construction rule, and fitness/cost function (chain length as a function of the parameters), drawing on 36 primary sources spanning the foundational number-theoretic literature (Brauer 1939 to Thurber's bounds) and the applied cryptographic-implementation literature (Gordon's 1998 survey onward). We complement the analysis with original empirical comparison. For accomplishing this we have implemented each parametrized method and measured actual chain lengths against an exact iterative-deepening ground truth across nine benchmark exponents. The results confirm the theoretically expected ordering (binary method the simplest but longest; sliding-window and factor method closest to optimal) and, notably, expose the real parameter-sensitivity of window-based methods -- a fixed window size k that performs well for cryptographically-sized exponents adds a precomputation overhead that is actively counterproductive for the small-to-moderate exponents common in comparative benchmarks, a practical caveat that is under-emphasized in much of the applied literature.