SV3.3B: A Sports Video Understanding Model for Action Recognition