The difference between the length of the central carbon–carbon bond and that of the adjacent flanked double bonds in polymers such as polyacetylene is closely related to their electronic properties and plays a central role in their conductivity upon doping. Simple as it seems, this bond length alternation (BLA) is a difficult test for many theoretical methods. Accurate coupled-cluster (CC) benchmark values are difficult to obtain even for small- and medium-sized oligoacetylenes due to their intrinsic computational limitations. Here we present a computationally much cheaper alternative to obtain accurate benchmark BLA values, even for large polyacetylene oligomers, by using the so-called spin-component scaled Møller–Plesset perturbation theory up to second order (SCS-MP2). Comparison between these new benchmark BLA with those provided by density functional theory (DFT) calculations shows a large dispersion of the results depending on the amount of exact exchange used in the exchange–correlation functional. We find that the percentage of exact exchange needed to accurately reproduce the new benchmark BLA is much larger than previously thought when comparison was made with values obtained using the MP2 method.