Data
Where event data stops being useful
Recruitment departments are buying certainty that the underlying data cannot supply. A short argument for narrower claims.
Athagora Research · 21 May 2026 · 7 min read
Event data has transformed how clubs describe players. It has been considerably less transformative in how clubs decide on them, and the gap between those two things is where expensive mistakes live.
Three limits recur. Sample size in lower-coverage leagues, collection inconsistency between providers and seasons, and the near-total absence of context for off-ball decisions that were correct but produced no event.
Cross-league translation is the most over-claimed application. Adjustment coefficients are estimated from the players who moved, and players who move are selected, the model is trained on survivors and asked to predict everyone.
None of this is an argument against analytics. It is an argument against the confidence interval being dropped somewhere between the analyst's notebook and the recruitment meeting.
The practical fix is procedural. Require every data-led recommendation to state the sample, the provider, the comparison set and the one thing that would most change the conclusion. Recommendations that cannot survive that format were never recommendations.
“The model is not wrong. The sentence written underneath it usually is.”