The Evidence on Teaching

The Relationship Between Deliberate Practice and Performance in Sports: A Meta-Analysis

Macnamara, B. N., Moreau, D., & Hambrick, D. Z. · 2016

grade Cmeta-analysisindependentreplicated
Population
Recreational to repeat Olympic gold medalists; youth and adult; team and individual, open- and closed-skill sports
Design
52 independent samples, 63 effect sizes, N=2,765 (38/63 published). Cross-sectional correlations of accumulated sport-specific practice with attained skill. Weighted mean accumulated practice 3,949 h (SD 2,942); subgroup means 4 h to 12,839 h. I2=83.5. Funnel symmetric; trim-and-fill estimated 0 missing. Skill-level moderator restricted to adult athletes (elite cell only 6 effect sizes, N=228).
Key findings
Practice explains 18% of sports performance variance overall but essentially nothing (1%, n.s.) among elite athletes — near the ceiling, hours no longer separate people. Earlier starting age did NOT predict higher skill (higher-skill athletes started 5.6 weeks later, n.s.), contradicting the early-specialization corollary. Restricting to solitary practice or objective performance measures did not rescue the effect; expert-rating outcomes gave the SMALLEST effect (2%), the opposite of what a measurement-artifact defence requires.
Genetic confound
None controlled. Elite-athlete status is ~30-80% heritable and VO2max TRAINABILITY itself ~47% heritable (HERITAGE), so the 82% unexplained plausibly contains both capacity and response-to-training genetics.
Replication notes
Consistent with the 2014 meta (sports 18%). Attacked by Ericsson (2016); the reply showed 20% using only objective measures, 22% using only individual practice, 17% excluding composites — robust to his objections.
DOI / URL
10.1177/1745691616635591

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Sports performance overallmeta-analytic r; % variancer=.43 [.35,.50]; 18% variance [12%,25%] — 82% unexplainedmixedattained expertise; cross-sectional; field/real-sportnonenot-applicabledomain-skill
Performance by skill level (the elite paradox)% variance (r)sub-elite 19% (r=.44, p<.01); mixed samples 29% (r=.54); ELITE only 1% (r=.11, p=.50, n.s.). Q(2)=7.04, p=.03mixedattained expertise; cross-sectional; fieldnonenot-applicabledomain-skill
Starting age of higher- vs lower-skill athletesmean difference in yearsHigher-skill started LATER by 0.11 yr (5.6 wk), [-0.41,0.62], p=.68 n.s. (24 effects, N=1,477). Starting-age difference did not predict practice difference (r=.18, p=.53)mixeddevelopmental/retrospective; fieldnonenot-applicabledomain-skill
Moderator: how performance was measured% variance (r)group membership 25% (r=.50); standardized objective 20% (.45); expert rating only 2% (r=.14). Q(2)=32.33, p<.001mixedattained expertise; fieldnonenot-applicabledomain-skill
Solitary deliberate practice onlyr; % variancer=.47 [.33,.58], 22% [11%,34%] — not different from non-solitary (18%), Q(1)=.25, p=.62mixedattained expertise; cross-sectional; fieldnonenot-applicabledomain-skill

Cited by