Coaches and Scouts

The 2020s brought computer vision and machine learning to video-based talent evaluation at a scale that genuinely threatens to replace some traditional scout functions. Systems like Synergy Sports (used across the NBA, NCAA, and other leagues) automatically tag and categorize every play in a game — half-court sets, transition plays, defensive schemes — without human review of each clip. BAM Sports and similar companies apply computer vision to soccer and basketball footage to extract player performance metrics automatically. Hudl has integrated AI tagging features that identify player locations, ball movement, and play types in uploaded video. In professional baseball, the combination of Statcast and machine learning has produced xFIP, xwOBA, and other "expected" statistics that strip luck from observed outcomes — metrics that some scouts now weight more heavily than traditional observation. The NHL deployed puck and player tracking across all arenas beginning with the 2019-20 season, the last of the four major North American leagues to implement full sensor tracking. The question the 2020s is raising: if a computer can watch 10,000 hours of prospect video and flag the top 50 candidates overnight, what is the value of a scout whose comparative advantage was watching those hours personally?

AI video analysis is compressing the pro scouting function at the margins — reducing the number of scouts needed for first-pass video screening while increasing demand for analysts who can interpret the outputs. But the broader coaching population (K-12, college, club sports) is largely insulated: the work of motivating athletes, designing practice, managing team chemistry, and making real-time decisions in competition is not yet a target for AI substitution. Frey & Osborne (2013) gave "Athletes, Coaches, Umpires" a low computerization probability (~0.18); Eloundou et al. (2023) similarly find low LLM exposure for coaching and athletic instruction tasks.