Forthcoming large-scale spectroscopic surveys will soon provide data on thousands of galaxy clusters. It is important that the systematics of the various mass estimation techniques is well understood and calibrated. We compare three different dynamical mass estimators using the C-EAGLE galaxy clusters, a set of high-resolution simulations with resolved galaxies a median total mass, M_{200c} = 10^{14.7} M_\odot. We quantify the bias and scatter of the Jeans, virial, and caustic mass estimators using all galaxies with a stellar mass M_*\gt 10^9 M_\odot, both in the ideal 3D case and in the more realistic projected case. On average we find our mass estimates are unbiased, though relative to the true mass within r200c the scatter is large with a range of 0.09-0.15 dex. We see a slight increase in the scatter when projecting the clusters. Selecting galaxies using the same criteria, we find no significant difference in the mass bias or scatter when comparing results from hydrodynamical and dark matter only simulations. However, selecting galaxies by stellar mass reduces the bias compared to selecting by total mass. Comparing X-ray derived hydrostatic and dynamical masses, the former are ˜30 per cent lower. We find a slight dependence between substructure, measured using two different metrics, and mass bias. In conclusion, we find that dynamical mass estimators, when averaged together, are unbiased with a scatter of 0.11 ± 0.02 dex when including interloper galaxies and with no prior knowledge of r200c.