Top panel: Variation of the median projected axis ratio bias with number of satellites. We compare the bias estimate from the MC with the actual bias from measuring shapes with satellites in the m₉ case. Here the bias is defined as (q_host−q_sat)/q_host where q_host is the axis ratio of the host DM halo and q_sat is that measured using the distribution of satellite galaxies. $b_{\mathrm{sat}}^{{\mathrm{m}}_9}$ denotes the bias when using the satellite positions to estimate the shape of the halo for the m₉ run, whereas $b_{\mathrm{sat}}^{{\mathrm{m}}_9}$ (BCG) denotes the case where the BCG is assumed to be the centre of the shape. 2p8 refers to the 2.8 Gpc box. The middle panel shows the deviation of the estimated bias from the MC exercise for the different hydrodynamic variations. Δ bias=b_MC−b_sat, and the error bars (shown by the shaded regions) are estimated using the jack-knife technique using 27 sub-boxes. The bottom panel shows the deviation from the MC exercise at different redshifts for the 2.8 Gpc run. The error bars are estimated by averaging results from eight sub-boxes.