This work studies the chemical kinetics of wood pyrolysis by comparing nine reduced reaction schemes against 22 microscale experiments of softwood and hardwood from the literature. The complexity of reaction schemes ranged from 1 to 12 reactions, with 2 to 7 species. Using multi-objective optimization for isothermal and non-isothermal conditions, the kinetic parameters for each reaction scheme were derived. It was found that the uncertainty of a prediction increases with the number of model parameters, but the accuracy does not always increase with the number of parameters. The appropriate reaction scheme for hardwood is three parallel reactions, as it presents the optimal balance between accuracy and uncertainty. For softwood, a higher complexity could be justified. This work shows the benefits of finding an appropriately complex kinetic scheme by building up complexity from simple schemes.