The Statistics That Changed How Teams Play
Traditional sporting statistics describe results: goals, runs, points. Analytics describes processes that reliably produce results — and the difference between the two has changed who gets signed, where, and for how much.
The measurement problem
A goal is a rare event, so it is a noisy measure of a player's contribution. A player who takes many high-quality shots may score fewer goals over a season than one who takes three bad ones at lucky moments. Analytics tries to measure the repeatable part — shot quality, field position, defensive pressure — rather than the noise at the end of it.
Expected goals and shot quality
In football and ice hockey, models estimate the probability that a given shot becomes a goal, based on location, angle, assistance and pressure. Summing those probabilities over many games predicts future scoring better than past goals do. That is why a striker on a long dry run can be a good signing: the underlying numbers say the chances are being created.
Where the limits are
Models inherit their assumptions. They tend to undervalue things that are hard to log — leadership, off-ball positioning, the effect of a player on teammates. And once an advantage is public, everyone copies it, so the edge decays. Analytics is a search process, not a finished answer.
Knowledge check
Two questions, no score
Why is a goal a noisy measure of performance?
Now prove it