Exceptional performance / A research essay
Before the
breakthrough.
How exceptional performers develop—and what we can reasonably know about their potential before the results arrive.
There is no verified formula for creating a champion. The more defensible picture combines developing skill, a suitable task, useful feedback and access to opportunities. Even then, competitive results depend on other people, resources and chance. Recognising potential means making better, revisable judgements—not discovering a permanent kind of person.
What builds capability?
Practice, knowledge, feedback and the conditions that let learning continue.
What forecasts success?
Information that improves a future forecast beyond a sensible baseline.
What changes the outcome?
Support whose effect can be distinguished from who received it.
A breakthrough rarely begins where the story begins.
In a January 2023 interview, the young jockey Billy Loughnane described a preparation that was already substantial: pony racing, work with coach Rodi Greene and rides supplied by his father’s stable. He also acknowledged the importance of good ponies and further opportunities after mistakes. Those are observations about a learning environment—not proof that one ingredient produced his rise.[1]
That dated account is useful because it precedes much of the career that an observer might now be tempted to read back into it. At that point, a reasonable judgement was that Loughnane warranted further opportunities and close observation. A forecast of sustained world-class performance would have required much more evidence.
The distinction matters in every field. We see a result; we infer an ability; then we invent a tidy story about how it formed. To resist that sequence, begin by defining the result we are trying to explain.
One excellent performance, repeated high performance, sustained world-class results and public recognition are different outcomes. So are capability and contribution. A useful forecast names the field, comparison population and time horizon: for example, performance among first-year apprentices next season, or a researcher’s contribution over the following decade.
Rankings create another complication. Better training might raise everyone’s absolute capability while leaving the number of champions unchanged. “This helps people improve” and “this helps us identify the eventual winner” are separate claims.
Four findings—and an important counterweight.
1. Practice matters, but hours are an incomplete explanation.
A widely discussed meta-analysis found that accumulated deliberate practice accounted for 18% of observed performance variation in sport and 21% in music. These are associations across people, not estimates that practice causes those percentages of anyone’s skill. Differences in measurement, opportunity and prior ability remain relevant.[6]
The definition is disputed. Ericsson and Harwell argued that broad measures omit essential features of expert-designed practice; Hambrick and colleagues challenged the narrower view’s empirical defensibility. The useful conclusion is to examine what practice contains, while avoiding a universal hours threshold or a precise partition of “talent versus effort”.[7] [8]
2. Feedback can improve learning. Its content matters.
Wisniewski and colleagues synthesised 435 educational studies involving more than 61,000 participants. After excluding extreme effects, the average effect was d = 0.48, with a reported interval of 0.44–0.51. Effects varied substantially. This supports attention to informative feedback; it does not show that a short coaching exercise identifies future champions.[9]
3. Early success and adult excellence are different targets.
In 51 study reports covering 6,096 athletes, Güllich and colleagues found different developmental patterns for high-performing juniors and world-class adults. The latter, compared with national-class adults, tended to have broader early sporting experience and slower initial progress. These histories are observational; they do not establish that delaying specialisation causes excellence.[2]
A separate review of longitudinal results from 13,392 athletes found a small pooled junior–senior performance correlation, r = 0.148. Age category mattered. This estimate concerns the included Olympic-sport populations and long horizons; it is not a universal measure of next-season predictability.[3]
A 2025 review extends the early-versus-adult distinction across sport, science, music and chess. However, limited overlap between two elite groups does not itself establish that early promise is useless—or harmful. A subsequent preprint raises precisely this base-rate and comparator problem. It is a methodological challenge, not a replication.[4] [5]
4. Opportunity helps make the record we later judge.
Randomised field experiments by van de Rijt and colleagues showed that small, arbitrarily assigned advantages could attract further rewards in online systems. That is causal evidence for accumulated advantage in those settings, not proof that recognition increases underlying skill.[10]
In Dutch research funding, applicants just above an early-career award threshold subsequently received more than twice the funding of applicants just below it over eight years, excluding the initial award. A near-identical starting assessment can therefore lead to a very different visible career record.[11]
The counterweight: some early measurements do carry information.
It would be a mistake to replace “spot the prodigy” with “early ability tells us nothing”. In a highly selected cohort of 563 young people, spatial ability measured at 13 added information beyond mathematical and verbal scores about creative output more than 30 years later. This is a domain-relevant predictive result, not a general-purpose test of future greatness.[15]
Evidence assessment: what the main studies can support
Evidence labels describe study design, not a universal quality ranking. No effect below is an estimated effect of the complete scouting protocol.
| Evidence | What was measured | What limits the inference |
|---|---|---|
| Practice Association | Performance differences correlated with accumulated practice; estimates differ by domain.[6] | Self-reported histories, selection and competing definitions complicate interpretation. Between-person variance does not identify the effect of adding an hour of training. |
| Feedback Intervention synthesis | 994 effects; the full pooled estimate was 0.55. Removing 35 extreme effects reduced it to 0.48.[9] | Residual heterogeneity remained high (I² = 83.4%). Treatments and outcomes differ; an average classroom-learning effect does not transfer automatically to elite careers. |
| Junior → senior Prediction | 129 effects; alternative dependence models gave pooled correlations of 0.145–0.151.[3] | Regional-to-international athletes, with Western European and male samples overrepresented. Withdrawals and late entrants may be missed; no full-population scouting calibration. |
| Initial rewards Randomised | Money, ratings, awards and endorsements were assigned experimentally.[10] | Online reward systems and short follow-up. Evidence for social feedback, not a lifetime effect on ability; larger initial rewards did not reliably create proportionately larger gains. |
| Dutch grants Natural experiment | Funding trajectories around a grant-score threshold.[11] | Local causal interpretation requires comparable applicants near the cutoff. Participation in later contests is part of the mechanism, not merely a nuisance variable. |
| NIH near misses Natural experiment | Early-career applicants near a funding threshold, followed for ten years.[12] | Attrition increased; stronger citation outcomes among those remaining require a separate interpretation from effects on everyone initially eligible. |
| Spatial ability Longitudinal | Later patents and publication categories; spatial measures added to earlier mathematical and verbal tests.[15] | Highly selected cohort. Incremental statistical association is not external temporal validation of an operational selection rule. |
Look beyond the familiar success story.
These seven comparisons combine three named trajectories with four cohort-level cases. The cohorts provide comparison groups that biographies usually lack. They are not seven equivalent experiments. “Knowable then” means what the audited records support at that stage, not an actual forecast made by this report.
| Case & observation point | Preparation, intervention & outcome | What was knowable beforehand | Rival explanation / boundary |
|---|---|---|---|
| Billy Loughnane January 2023 | Pony-racing preparation, coaching, family-stable rides and further chances after errors were already reported.[1] | Early professional competence and access to a supportive learning environment. Retained corrections were not independently measured. | Horse quality, apprentice allowances and ride allocation can alter results. Career treated as unfinished; no claim about ultimate peak. |
| Tiger Woods By the end of 1996 | Three consecutive U.S. Junior Amateur titles (1991–93), then three U.S. Amateur titles (1994–96), preceded U.S. Open wins in 2000, 2002 and 2008.[16] | Repeated, publicly recorded performance at progressively demanding competition levels. The audited record does not isolate a particular coaching intervention. | Shows that early excellence can persist. One selected champion cannot estimate the conversion rate of all outstanding juniors or establish an optimal childhood programme. |
| Katalin Karikó 1995 / 2005 | Her later account describes demotion in 1995 while retaining bench access. Collaboration with Drew Weissman produced the pivotal 2005 work; both received the 2023 Nobel Prize.[17] [18] [19] | In 1995: technical experience and a research programme, without a demonstrated vaccine platform. In 2005: a published mechanistic result worth testing and extending. | Later recognition depended on complementary discoveries, delivery technology and collaborators. Persistence alone does not explain the outcome; adversity was not experimentally beneficial. |
| Junior athletes Youth → open-age sport | The longitudinal synthesis finds weak average continuity of relative performance across these stages.[3] | A current ranking was observable, but its long-horizon value depended on age and comparison group. No common intervention was assigned. | Changing maturation, competition and participation matter. A lower adult rank, injury, voluntary exit and missing follow-up must not all be labelled “failure”. |
| Talented adolescents Age 13 → midlife | Among 563 participants already selected for high ability, spatial assessments added predictive information about later creative-output categories.[15] | Specific assessed strengths, before the outcomes existed. This was not an experiment in increasing spatial ability. | Challenges an explanation based only on practice or access. Generalisation to the full population and to exceptional prizes is unestablished. |
| Dutch grant applicants Either side of the cutoff | Award winners and near-miss applicants had similar review scores but subsequently accumulated different funding records.[11] | Proposal assessments and threshold position. The award itself changed resources and potentially later selection. | Large support differences did not establish corresponding differences in scientific capability. Funding, publication impact and contribution are separate outcomes. |
| NIH grant applicants Early-career near miss / narrow win | Near misses were more likely to leave the NIH system. Among those remaining, later publication impact exceeded that of narrow winners.[12] | Similar proposal scores near the funding threshold; later survival was not knowable at selection. | A challenge to a simple “more initial support always means higher impact” story. Leaving this system need not mean leaving science. Survivor comparisons cannot justify manufacturing setbacks. |
Comparison coverage: sustained early excellence; uncertain ongoing development; delayed recognition; restricted opportunity; imperfect continuity of early promise; and support that changes status more clearly than demonstrated capability.
Karikó’s story is especially easy to oversimplify. The documented scientific pathway contains specific experimental problems and complementary expertise. Her eventual award should not become a claim that stubbornness predicts a Nobel Prize, or that scarce resources help people discover more.[17] [19]
Capability and opportunity grow together.
The simplest useful model is a process, not a weighted equation. Existing abilities and experience influence what someone can learn. Practice and feedback can change capability. Tasks, equipment, collaborators and competitive conditions influence how that capability becomes a result. Results then affect reputation and access to the next opportunity.
+ useful feedback
capability
in real conditions
Starting abilities and experience influence learning and performance. Resources and task fit influence practice and results. Opportunity also affects results directly. Health, recovery and engagement can sustain or interrupt participation; competitive conditions and chance affect the observed outcome.
This diagram distinguishes capability, a possible mediator of training, from results, an outcome, and selection, a mechanism that changes subsequent exposure. It also exposes confounding: access to a better horse may raise a jockey’s win rate without an equivalent improvement in riding skill.
If estimating the total effect of an early grant, automatically adjusting away the later opportunities created by that grant would remove part of the pathway of interest. Conversely, if trying to measure skill independently of opportunity, task difficulty and resources must be recorded. The adjustment depends on the question.
Yin and colleagues’ work on repeated attempts offers a related hypothesis: the way attempts change may be more informative than their sheer number. Their model and observational sequences motivate measuring improvement, but do not validate “learns from mistakes” as a general scouting test. The paper’s 2020 correction concerns axis markers and labels in Figure 3d–f.[13] [14]
Where the model needs different settings
Structured performance
In tasks with reliable feedback and repeatable conditions, assess technical errors and retention over multiple sessions. Increase difficulty while recording equipment, opposition and fatigue. Safety-critical errors require qualified oversight, not an unconditional promise of another attempt.
Open-ended discovery
In science, feedback is delayed and a failed result may still be informative. Review experimental choices, reproducibility and responses to disconfirming evidence. Use longer horizons and assess contributions separately from grants, journal prestige and awards.
These are reasoned applications of the model, not tested cross-domain prescriptions. Health, recovery, family support and sustained engagement belong in the model as possible enabling conditions; this review does not estimate their separate causal weights.
Five tensions that should remain tensions
| Tension | Practical interpretation to test |
|---|---|
| Specialisation / breadth | Use breadth to explore fit and acquire varied experience; specialise as domain demands and informed preferences become clearer. No universal starting age follows from group averages.[2] |
| Persistence / redirection | Continue when there is retained learning or a plausible removable constraint. Change task, method or objective when repeated work adds no information and costs keep rising. |
| Coachability / independence | Look for accurate understanding and evidence-based adaptation, including justified disagreement. Compliance is not the same as learning. |
| Fast progress / durable progress | Use delayed tests and new conditions to distinguish an immediate gain from something that lasts. A fast start can reflect prior exposure; a slow start can reflect an unfamiliar task. |
| Confidence / calibration | Ask for a forecast before the attempt and an explanation afterwards. Compare predictions with results rather than rewarding confident presentation. |
Except where cited, these are proposed decision principles. None is a validated personality test.
Observe the next learning cycle.
A practical approach begins with ordinary performance measures and then asks whether repeated observation adds useful information. “Winner mentality” is too vague to audit. A change visible in a dated work sample can be checked.
Proposed / unvalidated The following protocol is a candidate assessment process. It is suitable for a pilot, not for confident claims about anyone’s eventual ceiling.
- Set the decision and its horizon.Specify the task, outcome, comparison group and review date. Separate selecting the strongest current candidate from allocating support where it may have the greatest effect.
- Collect a representative starting sample.Use more than one task or occasion. Record current performance, age or career stage, experience, earlier trajectory, resources, task difficulty and available opportunities. Establish measurement reliability.
- Provide a clear, credible correction.Name an observable error and a practicable adjustment. Record the instruction and whether the participant understood it. Make instruction time and task access comparable across candidates.
- Test retention and transfer.After an appropriate delay, repeat an equivalent task without prompting. Then use an unfamiliar version that calls for the same principle. Choose the delay for the domain before seeing results.
- Increase challenge and test a constraint.Raise difficulty gradually. Where feasible, offer a time-limited change in coaching, access or equipment to examine an identified obstacle. Keep the original performance record and the support record separate.
- Review, revise and preserve re-entry.Use evidence of learning to continue support; revise the method when gains fail to persist; change environment when a constraint dominates. Reconsider the objective when its costs, fit or the participant’s preferences change. Record uncertainty and a date for reassessment.
| Candidate signal | Observation, timing & comparison | False positive / false negative | Evidence & practical test |
|---|---|---|---|
| Retained correction | The same material error becomes less frequent on a later unprompted task, compared with people at similar baseline and exposure. | Positive: rehearsing a known test. Negative: fatigue or misunderstood instructions suppressing a real gain. | Feedback evidence motivates the test; incremental forecasting value is unknown. Blindly rate equivalent tasks before and after a delay.[9] |
| Transfer | A learned principle works on an unfamiliar task of comparable difficulty, after a learning block. | Positive: the “new” task was already familiar. Negative: extra knowledge demands conceal transferable skill. | Mechanism-based hypothesis. Audit prior exposure, vary surface features and test whether the gain predicts later independent work. |
| Adaptation to difficulty | Choices and error patterns improve after a defined increase in challenge; compare candidates with similar starting competence and opportunity. | Positive: easier assignments or better equipment. Negative: ceiling effects, illness or adjustment costs. | Attempt-dynamics research is indirect support. Model difficulty explicitly and compare with current performance plus past trajectory.[13] |
| Judgement & self-correction | Before and after attempts, the person identifies consequential errors, forecasts likely results and revises an explanation when contradicted. | Positive: fluent retrospective explanation. Negative: strong practical judgement expressed poorly in interviews. | Unvalidated here. Capture forecasts before the outcome; score decisions and work products separately from verbal polish. |
| Performance given opportunity | Repeated work is assessed alongside coaching, access, equipment and task allocation at the prediction date. | Positive: hardship mistaken for hidden ability. Negative: missing records conceal preparation or constraints. | Opportunity studies justify measuring exposure, not an automatic “disadvantage bonus”. Test matched access or a defined support offer.[10] [11] |
| Sustained engagement | Over several cycles, the person follows through, seeks useful information and adjusts effort appropriately; compare with realistic time and access constraints. | Positive: attendance without learning. Negative: caring duties or finances reduce visible availability. | A practical feasibility signal; no independent predictive weight is established here. Log opportunities offered, attendance, reasons for absence and learning separately. |
Who is likely to succeed under current conditions? This is a forecasting problem. Who would improve most with additional support? This is a treatment-effect problem. A high forecast does not imply a large benefit from support, and a low forecast does not imply a small benefit. Observational “potential scores” cannot resolve the difference.
Do not add the indicators into a numerical “potential score” without evidence for the weights. During a pilot, retain the individual observations, disagreements between raters and reasons for decisions. A short assessment should open a development conversation, not close a career.
A good story still has to beat a baseline.
No suitable, ready-to-use public dataset was identified in this review that jointly measures these candidate indicators, prior opportunity and later outcomes. No forecasting model was fitted. A prospective study is the defensible next step.
A concrete prospective design
Start within one field and developmental stage—for example, new entrants to several comparable sports programmes. Register a 12-month continuous performance outcome and a separate, later adult-performance outcome. The latter needs its own definition and enough years of follow-up; the first cannot stand in for it.
Enrol successive intake cohorts, including people who are not selected for intensive support. Use the first cohort to develop the measures and a restrained model; lock both before evaluating a later cohort. Keep every observation from one person in the same partition. Add a different programme as an external test if access permits.
Compare three forecasts: current performance alone; current performance plus age or stage, experience, prior trajectory and resources; and that stronger baseline plus the candidate indicators. Evaluate the added indicators in the same held-out people. A more elaborate method earns its place only through a useful, reproducible improvement.
Choose the sample size by simulation using the expected outcome rate, measurement reliability, clustering and the smallest worthwhile improvement. If exceptional outcomes are too rare to support precise evaluation, extend recruitment or follow-up. Do not manufacture certainty from a handful of successes.
Measurement, error and decision rules for the study
- Prevent leakage: timestamp all inputs; use only records available at the forecast date. Blind outcome raters to early labels and keep later reputation out of baseline features.
- Measure reliability: use repeated tasks, blinded duplicate ratings and a documented rubric. Inspect disagreement and floor or ceiling effects before claiming stable differences.
- Track everybody: distinguish injury, voluntary exit, career changes and lost contact. Treat ongoing careers as censored at the evidence cutoff. Report missingness by group and use sensitivity bounds; do not silently call missing outcomes failures.
- Handle dependence: account for repeated observations and clustering within coaches, programmes and cohorts. Evaluate temporal drift when conditions change.
- Report useful performance: for continuous outcomes, report prediction error and improvement over baseline with uncertainty. For binary outcomes, report precision, recall, a confusion matrix, precision–recall performance, and probability calibration, including Brier score and calibration plots.
- Price the errors: report support cost, false positives and missed high performers at the actual capacity limit. Compare with selecting on baseline performance and, where appropriate, a lottery among similarly qualified candidates.
- Test incremental value: predefine the minimum useful gain. If added signals fail to beat the strong baseline in later cohorts, drop or revise them. Avoid post hoc thresholds and subgroup claims.
- Separate the intervention: where feasible, randomise a defined support package among equally eligible candidates, analyse all assigned participants and report uptake. This estimates the support offer’s effect; individual benefit rankings need further validation.
Why rarity changes the meaning of a “good” test
Consider a purely illustrative group of 10,000 candidates, of whom 1% will achieve the specified exceptional outcome. A test that catches 80% of eventual achievers, while incorrectly flagging 10% of everyone else, identifies 80 true positives and 990 false positives.
| Later outcome | Flagged | Not flagged | Total |
|---|---|---|---|
| Exceptional outcome | 80 | 20 | 100 |
| Other outcome | 990 | 8,910 | 9,900 |
| Total | 1,070 | 8,930 | 10,000 |
Only 7.5% of those flagged become exceptional in this example. That is still a considerable improvement over the 1% starting rate. A useful forecast can therefore have many false positives; whether it is worth using depends on the cost of support and the cost of missing people.
Reproduce the base-rate calculation
Achievers = 10,000 × 0.01 = 100
True positives = 100 × 0.80 = 80
False positives = 9,900 × 0.10 = 990
Precision = 80 ÷ (80 + 990) = 0.074766… ≈ 7.5%
False negatives = 100 − 80 = 20
All inputs are hypothetical. These numbers are a teaching example, not measured accuracy for the proposed protocol. No study effect size was converted into a success probability.
Keep the model open to an inconvenient result.
The strongest challenges to a simple learning-and-opportunity account are already visible. Specific early abilities can carry long-term information. Some early champions remain champions. Initial support can strongly affect later funding without proving a corresponding improvement in scientific contribution. Near misses can show stronger later impact among survivors even while increasing attrition.[11] [12] [15] [16]
These findings argue for a model with multiple pathways. They also limit the claim that one behavioural pattern will travel intact from jockeys to researchers. Fast feedback, physical maturation, team dependence, long experimental delays and scarcity of recognition create different selection problems.
If retained correction and transfer add no useful predictive information beyond baseline ability, experience and resources in later cohorts, their scouting role should be rejected. If a well-implemented support package improves neither delayed learning nor subsequent performance, the proposed intervention needs revision. If gains appear only on rehearsed tasks, the claim of usable capability is too strong.
Avoid five seductive narratives: the inevitable prodigy; the magic practice total; the heroic survivor of adversity; the coach’s instinct vindicated by one success; and the late bloomer whose later achievement is treated as obvious in retrospect. Each story hides a comparison group.
The practical aim is narrower and more useful: create conditions in which capability can emerge, observe development fairly, and maintain forecasts that can change. Some uncertainty about exceptional achievement will remain even when the learning environment is excellent.
Bibliography
Numbered citations in the article link directly to the sources below. Accessed 20 October 2026. Publication years distinguish online-first dates from issue dates where relevant. The list includes primary research, scholarly reviews, official records and clearly labelled testimony.
- Contemporaneous interview
Nordoff, J. (2023). Interview with trainer Mark Loughnane and star apprentice Billy Loughnane. Sporting Life, 31 January.A dated account of preparation, coaching and access; testimony does not isolate their effects.https://www.sportinglife.com/racing/news/interview-with-trainer-mark-loughnane-and-star-apprentice-billy-loughnane/207009 - Meta-analysis; association
Güllich, A., Macnamara, B. N., & Hambrick, D. Z. (2022; online 2021). What Makes a Champion? Early Multidisciplinary Practice, Not Early Specialization, Predicts World-Class Performance. Perspectives on Psychological Science, 17(1), 6–29.Developmental histories of 6,096 athletes. Group differences do not establish an optimal training prescription.https://doi.org/10.1177/1745691620974772 - Longitudinal predictive synthesis
Barth, M., Güllich, A., Macnamara, B. N., & Hambrick, D. Z. (2024; online 2023). Quantifying the Extent to Which Junior Performance Predicts Senior Performance in Olympic Sports: A Systematic Review and Meta-analysis. Sports Medicine, 54, 95–104.Junior-to-senior associations; populations and attrition limit extrapolation.https://doi.org/10.1007/s40279-023-01906-0 - Cross-domain review
Güllich, A., Barth, M., Hambrick, D. Z., & Macnamara, B. N. (2025). Recent discoveries on the acquisition of the highest levels of human performance. Science, 390(6779), eadt7790.Publisher-indexed abstract checked. Included as an extension of the debate, not independent causal confirmation.https://doi.org/10.1126/science.adt7790 - Unreviewed methodological commentary
Xiao, W., Zhang, X., & Xiao, W. (2026; manuscript dated 2025). Commentary on “Recent Discoveries on the Acquisition of the Highest Levels of Human Performance”: A Critique of the Base-Rate Fallacy, Logical Paradox, and the “Label-Sorting” Paradigm. SSRN preprint, posted 21 January.Abstract checked; raises comparator and base-rate concerns. Not a replication or a settled rebuttal.https://ssrn.com/abstract=5980735 - Meta-analysis; association
Macnamara, B. N., Hambrick, D. Z., & Oswald, F. L. (2014). Deliberate practice and performance in music, games, sports, education, and professions: a meta-analysis. Psychological Science, 25(8), 1608–1618.Variance explained differs by domain; the estimate is not the causal share of an individual's achievement.https://doi.org/10.1177/0956797614535810 - Methodological critique
Ericsson, K. A., & Harwell, K. W. (2019). Deliberate Practice and Proposed Limits on the Effects of Practice on the Acquisition of Expert Performance: Why the Original Definition Matters and Recommendations for Future Research. Frontiers in Psychology, 10, 2396.Challenges broad operational definitions of deliberate practice.https://doi.org/10.3389/fpsyg.2019.02396 - Review and response
Hambrick, D. Z., Macnamara, B. N., & Oswald, F. L. (2020). Is the Deliberate Practice View Defensible? A Review of Evidence and Discussion of Issues. Frontiers in Psychology, 11, 1134.The opposing assessment of definitions and testability in the practice debate.https://doi.org/10.3389/fpsyg.2020.01134 - Intervention meta-analysis
Wisniewski, B., Zierer, K., & Hattie, J. (2020). The Power of Feedback Revisited: A Meta-Analysis of Educational Feedback Research. Frontiers in Psychology, 10, 3087.Strong heterogeneity; educational learning outcomes do not establish elite-career forecasting validity.https://doi.org/10.3389/fpsyg.2019.03087 - Randomized field experiments
van de Rijt, A., Kang, S. M., Restivo, M., & Patil, A. (2014). Field experiments of success-breeds-success dynamics. Proceedings of the National Academy of Sciences, 111(19), 6934–6939.Arbitrary initial advantages affected subsequent rewards in online social systems.https://doi.org/10.1073/pnas.1316836111 - Regression discontinuity
Bol, T., de Vaan, M., & van de Rijt, A. (2018). The Matthew effect in science funding. Proceedings of the National Academy of Sciences, 115(19), 4887–4890.Local comparison around a Dutch grant threshold; later funding is distinct from underlying capability.https://doi.org/10.1073/pnas.1719557115 - Regression discontinuity and longitudinal analysis
Wang, Y., Jones, B. F., & Wang, D. (2019). Early-career setback and future career impact. Nature Communications, 10, 4331.Setbacks increased attrition; favourable impact among survivors is not an argument for withholding support.https://doi.org/10.1038/s41467-019-12189-3 - Observational sequences and process model
Yin, Y., Wang, Y., Evans, J. A., & Wang, D. (2019). Quantifying the dynamics of failure across science, startups and security. Nature, 575, 190–194.Suggests informative attempt dynamics; does not validate the scouting protocol proposed here.https://doi.org/10.1038/s41586-019-1725-y - Correction
Yin, Y., Wang, Y., Evans, J. A., & Wang, D. (2020). Author Correction: Quantifying the dynamics of failure across science, startups and security. Nature, 582, E16.Corrects misplaced axis markers and omitted labels in Figure 3d–f; checked alongside the original paper.https://doi.org/10.1038/s41586-020-2325-6 - Longitudinal predictive study
Kell, H. J., Lubinski, D., Benbow, C. P., & Steiger, J. H. (2013). Creativity and technical innovation: Spatial ability’s unique role. Psychological Science, 24(9), 1831–1836.Specific early abilities added information in a highly selected cohort; not a universal talent test.https://doi.org/10.1177/0956797613478615 - Official competition record
United States Golf Association (Undated; accessed 2026). U.S. Junior Amateur Records. USGA official records.Confirms Woods’s junior, amateur and U.S. Open titles; not an evaluation of his coaching.https://www.usga.org/content/usga/home-page/media/online-media-center/usga-records/u-s–junior-amateur-records.html - Scientific historical account
Nobel Assembly at Karolinska Institutet (2023). The Nobel Prize in Physiology or Medicine 2023: Advanced information. NobelPrize.org.Documents the experimental and collaborative pathway to modified-mRNA discoveries.https://www.nobelprize.org/prizes/medicine/2023/advanced-information/ - Retrospective first-person testimony
Nobel Prize Outreach (2024). Katalin Karikó: Nobel Prize Conversations. NobelPrize.org podcast and transcript.Her account of demotion and continued access to a research bench; subject to hindsight.https://www.nobelprize.org/prizes/medicine/2023/kariko/podcast/ - Official award announcement
Nobel Assembly at Karolinska Institutet (2023). Press release: The Nobel Prize in Physiology or Medicine 2023. NobelPrize.org, 2 October.Confirms the shared prize and the chronology of foundational discoveries.https://www.nobelprize.org/uploads/2023/10/press-medicineprize2023-3.pdf
Scope, verification & return manifest
What was reviewed—and what was not
This is a targeted research synthesis based on Exceptional Performance Research Charter, supplied by the user, with a source cutoff of 20 October 2026. Searches began with the charter’s four starter sources and expanded to practice, feedback, junior-to-senior prediction, cumulative advantage, longitudinal ability studies and contrasting funding outcomes.
Primary papers and official records were prioritised. Full accessible text was examined where available; several paywalled studies and the 2025 review were assessed through publisher or PubMed abstracts. That limits independent scrutiny of detailed analyses. This is not an exhaustive systematic review and has no claim to complete coverage of the literature.
Source dates and study designs were checked. The junior–senior result was checked against the original methods and limitations; feedback estimates were checked to distinguish the full estimate from the outlier-adjusted estimate. Biographical facts were tied to dated interviews or official records. Record access and abstract availability should not be mistaken for an independent reanalysis.
The Yin et al. correction was located and its axis-label changes checked. The practice-definition dispute and a 2026 preprint critique of the 2025 review were included. No replication of the complete proposed framework was identified. Search access failures on some sites were addressed through indexed publisher text, PubMed or official PDF versions.
Requirements covered, deviations and unresolved questions
- Direct answer and evidence assessment: opening synthesis, four findings, counterweight and study-design table.
- Comparison set: seven cases, comprising three individuals and four cohorts. Cohorts replace four additional biographies to expose denominators and selection effects. Detailed training records were not available for every individual; these gaps are marked rather than reconstructed speculatively.
- Model and boundary conditions: conceptual pathways, rival causes, domain adaptations and five tensions. No estimated weights or universal thresholds.
- Scouting and intervention: observable indicators, timing, comparison groups, false positives, false negatives and repeatable pilot procedures. Decision principles are proposals, not empirically optimised rules.
- Validation: prospective temporal and external testing, strong baselines, measurement checks, calibration and attrition handling. No bounded empirical test was performed because the required linked observations were not available in a ready-to-use public dataset identified here.
- Reproducibility: the only original numerical calculation is the explicitly hypothetical base-rate example. Formulae, inputs and a machine-readable calculation record are included in the download package.
- Unresolved: whether retained learning adds value beyond ability and exposure; who benefits most from extra support; how reliably delayed contribution can be measured; and which mechanisms transfer between domains.
- Scope boundary: sport and science carry the argument; music provides a limited comparison from meta-analysis. No exhaustive treatment of entrepreneurship, genetic prediction or clinical interventions is attempted.
Files, image credits and offline use
Article: exceptional-performance.html is self-contained. The two illustrations, layout and navigation are embedded; no connection is required to read it. Bibliography destinations require internet access.
Companion package: the HTML, two original PNG illustrations, bibliography.json, calculation.json and image-credits.txt. The credit file records both complete generation prompts and the built-in image generation method.
Both pictures are AI-generated editorial illustrations created for this article with the built-in imagegen tool. They depict fictional scenes, not documentary photographs of the named people. They do not serve as research evidence. The process diagram is an original conceptual illustration.
Critical dependencies: the source literature for substantive findings and future access to prospectively collected data for validation. No paid database, new experiment or independent dataset replication was used.
Prepared with AI assistance. The evidence labels and explicit limits are part of the article’s conclusions. This report does not assign a probability of future greatness to any named individual.
↑ Back to the beginning