Marker-based inverse kinematics (IK) is prone to errors arising from measurementnoise and soft-tissue artefacts. Various least-squares and Bayesian methods canbe applied to limit the estimation error to a minimum. Recently proposed meth-ods like Bayesian IK come at an increased computational cost however. In thistechnical paper, we present an overview of eight different least squares or BayesianIK methods, including their accuracy and computational load for IK problemsinvolving a single rigid body and three rotational degrees-of-freedom, whose at-titude is estimated from four noisy marker positions. The results indicate thatNon-Linear Least Squares, Variational Bayesian and full Bayesian IK are supe-rior to Singular Value Decomposition in terms of accuracy, with approximatelya two-fold error reduction. However, only Non-Linear Least Squares and Varia-tional Bayesian IK are computationally efficient enough to scale towards practicaluse in biomechanical applications, with computational durations of 1-10 ms; fullyBayesian procedures required approximately 30 s for single rotation calculations.All Python code and supplementary material can be found in this paper’s GitHubrepository: https://github.com/benserrien/pybik.