Real-time ship motion predictions contributes to safer operations at sea and increases workability. Nowadays, a handful of people and companies are actively working on this subject. Most approaches use the ship’s navigational radar to first predict the wavefield surrounding the vessel and then calculate the motion response. One of the methods used to calculate the ship’s motions is a linear ship motion model based on a frequency-domain approach. The downside of this approach is that the accuracy of the motion predictions is affected by (for example) uncertainties in the Response Amplitude Operators. In this thesis, it is shown that estimating transfer-functions from measured motions and a so-called "now-cast" prediction of the forces will counteract for such uncertainties. For the estimation of these transfer-functions, different methods and smoothing techniques are evaluated. Based on sea-trial data, it is shown that the accuracy of the motion predictions increases with ~ 1-10 % when estimated transfer-functions are used, compared to the solutions obtained by pre-calculated transfer-functions.