Lessons from a blind study of simulated lenses:image reconstructions do not always reproduce trueconvergence

In the coming years, strong gravitational lens discoveries are expected to increase infrequency by two orders of magnitude. Lens-modelling techniques are being developedto prepare for the coming massive influx of new lens data, and blind tests of lensreconstruction with simulated data are needed for validation. In this paper we presenta systematic blind study of a sample of 15 simulated strong gravitational lenses fromthe EAGLE suite of hydrodynamic simulations. We model these lenses with a free-formtechnique and evaluate reconstructed mass distributions using criteria based on shape,orientation, and lensed image reconstruction. Especially useful is a lensing analogue ofthe Roche potential in binary star systems, which we call thelensing Roche potential.This we introduce in order to factor out the well-known problem of steepness or mass-sheet degeneracy. Einstein radii are on average well recovered with a relative error of∼5% for quads and∼25% for doubles; the position angle of ellipticity is on averagealso reproduced well up to±10◦, but the reconstructed mass maps tend to be tooround and too shallow. It is also easy to reproduce the lensed images, but optimisingon this criterion does not guarantee better reconstruction of the mass distributio