Quantitative agreement between experimentally determined M-C bond dissociation energies (BDE) and DFT predictions (M06-2X/TZ//TPSSTPSS/DZ) can be reached by choosing the correct anchor for experimentally derived BDE. For the example of the archetypical metallocene catalyst Cp2TiCl2, it is shown that titanium-carbon bonds are very weak under polymerization conditions and fluctuate; steric strain is introduced after 2,1 insertion and via olefin capture. Thus, homolysis can become competitive with chain propagation. Depending on the catalyst and temperature, 2,1 insertion can be only a temporary inconvenience (dormancy) or a definitive decay event. It is then shown for a set of nine common Ti and Zr polymerization catalysts how ligand variation affects the metal-carbon BDE. Predicted stabilities of the M(IV) oxidation state with respect to homolysis are in nice agreement with the experimentally observed temperature tolerance of the various catalysts: homolysis is easier for Ti than for Zr, and cyclopentadienyl groups in particular facilitate homolysis, especially in bis-cyclopentadienyl systems. (C) 2017 Elsevier Inc. All rights reserved.