In recent years, Large-Eddy Simulation (LES) is spotlighted as an engineering tool and severe research efforts are carried out on its applicability to Internal Combustion Engines (ICEs). However, there is a general lack of definitive conclusions on LES quality criteria for ICE. This paper focuses on the application of LES quality criteria to ICE and to their correlation, in order to draw a solid background on future LES quality assessments for ICE. In this paper, TCC-III single-cylinder optical engine from University of Michigan is investigated and the analysis is conducted under motored condition. LES quality is mainly affected by grid size and type, sub-grid scale (SGS) model, numeric schemes. In this study, the same grid size and type are used in order to focus on the effect on LES quality of SGS models and blending factors of numeric scheme only. In the first section of the study, single grid estimators are used to compare two sub-filter models which are static Smagorinsky model and dynamic Smagorinsky model. Also, two cases which are simulated with different blending factors for numeric schemes and same SGS model are compared. In the second section, the in-cylinder gas-dynamics and flow structures are analyzed by comparing experimental results (pressure transducers and Particle Image Velocimetry (PIV) velocity fields) with a dataset of consecutive LES cycles. The flow analysis focuses at four different crank angle positions (bottom dead center (BDC), middle of exhaust and intake valve opening timing and mid-compression stroke) on the same section plane as PIV visualizations. Finally, the connection between the LES quality criteria and the accuracy of simulation results with experiments is discussed and conclusions are drawn to outline a best practice in LES quality for ICE.