ATLAS top modeling: NLO subtraction and PS uncertainties

A new generation of top quark pair Monte Carlo samples with modified tuning and settings is now available. These new samples are used along with an updated prescription for evaluating the top quark modelling uncertainty, with results compared with previous uncertainty prescriptions. Different ways of evaluating the matching uncertainty are compared to find an approach that avoids the overlap of different effects. The new prescription also revisits the estimation of the parton shower uncertainty and uses a new breakdown of the ISR uncertainty into several components, namely variations in the scale and $h_{damp}$ parameters and the shower tune. Moreover, to facilitate the evaluation of the uncertainty through internal variations of a single generator, studies are performed using the \SHERPAV{2.2.8} generator. The uncertainties quoted in these figures are calculated as the difference between generators at particle level, and don't directly correspond to the typical modelling uncertainties used in ATLAS measurements as detector-level effects are omitted. Nevertheless, the comparisons shown in these plots provide a good indication on the performance that could be achieved using the new MC samples. Observables measured by the ATLAS collaboration in three top-quark pair decay channels are considered in this study.