It would be quite difficult to make it look like the real thing at this stage. Also, you will need to be creative in many areas, for example, in the way you link the video to the sound of the play-by-play announcer.
You may be able to find models and programs in Spaces that you can use as components.
My personal concern is that I have not seen a solution suitable for generating long videos with a storyline.
The current generation AIs are good for short videos, audio and still images respectively, but the best practices for combining them all are all still being explored.