When Does Expert Intuition Beat Statistical Models?

## Research Question When does expert intuition beat statistical models in sports decision-making? What conditions make a coach's gut feel more reliable than analytics? ## Key Findings ### Conditions for Trustworthy Intuition From "Integrating Intuition and Information for the Development of Coaching Expertise" (doi:10.51224/srxiv.588): - **High validity environment:** The setting has valid, specifiable cues with a predictable relationship to outcomes - **Adequate opportunity:** Prolonged, repeated exposure to those cues with feedback on outcomes - **Closed-skill individual sports** (swimming, track) meet these conditions better than open-skill team sports - **Risk:** Intuition is prone to representativeness and availability biases, plus an "illusion of validity" in unfamiliar contexts ### Swimming Coach Study (Crowcroft et al., 2020) - Nine elite swimmers, 93 races - Coach with 13 years' experience predicted race outcomes without monitoring data - Described as "high accuracy" by observational analysis - Subjective fatigue reports and morning HRV were NOT reliably associated with performance changes - **Limitation:** Single small study; individual sport; may not generalize to team sports - Source: https://www.scienceforsport.com/can-a-coach-rely-on-intuition-alone/ ### Team-Sport Evidence (Cuts Against Intuition) - O'Brien-Smith et al.: Coaches judging footballers' cooperative behavior were consistently misaligned with objective measures regardless of experience level - Source: https://doi.org/10.51224/srxiv.588 ### Intuition Plus Data (Taylor & Francis, 2025) - Sparse, interpretable models combined with domain expertise often perform competitively with black-box models - Human judgments can carry information not in the metrics - Expert adjustment of model output works best when the expert knows what cues the model uses - Overreliance on model output is a risk under high stakes and uncertainty - Source: https://www.tandfonline.com/doi/full/10.1080/24733938.2025.2517056 ### Talent Selection Study - "Science or Coaches' Eye? – Both!" (PMC6370964): Combining multidimensional measurements with coach assessments was fruitful for youth football talent selection - Source: https://pmc.ncbi.nlm.nih.gov/articles/PMC6370964/ ### Coaches' Self-Reported Intuition - Qualitative study of 18 high-level British coaches: reported high use of intuitive decision-making, with experience as precursor - Pattern: intuitive first, then deliberate checking - **Limitation:** Self-report does not establish accuracy - Source: https://pmc.ncbi.nlm.nih.gov/articles/PMC4830814/ ## Implications for the Counter-Example Search The evidence suggests expert intuition is most trustworthy in: 1. Stable, predictable environments with clear cue-outcome relationships 2. Individual/closed-skill sports where the expert gets repeated feedback 3. Settings where the expert knows what the model uses and can identify what it misses Team sports like NFL, NBA, MLB, NHL, and soccer are **open-skill, high-variability environments** where intuition is more susceptible to bias. This may explain why verified counter-examples are absent: the conditions that would make a coach's gut demonstrably better than a model simply don't exist in these sports. ## Key Sources - https://doi.org/10.51224/srxiv.588 (Integrating Intuition and Information) - https://www.scienceforsport.com/can-a-coach-rely-on-intuition-alone/ (Crowcroft swimming study) - https://www.tandfonline.com/doi/full/10.1080/24733938.2025.2517056 (Taylor & Francis football analytics) - https://pmc.ncbi.nlm.nih.gov/articles/PMC6370964/ (Science or Coaches' Eye) - https://pmc.ncbi.nlm.nih.gov/articles/PMC4830814/ (Intuitive Decision Making)