Research Papers

Asad Khan, Arpan Bhattacharya

“Fast Hybrid Search for Red-Zone Play Recommendations” treats NFL red-zone playcalling as an information-retrieval problem: given the current game situation and recent play context, it searches a large corpus of historical red-zone snaps to surface similar precedents a coach might recall from film. Each play is converted into a compact, coach-like text “play card,” and a hybrid search stack combines keyword matching with semantic embeddings to retrieve candidates quickly. A constrained lightweight language model then synthesizes those candidates into a concise “play family” recommendation that fits existing coaching workflows and can be linked back to real historical analogues.