The evidence room
Keep asking the same three questions.
What did they bring? What were they handed? What surrounded them? Then inspect the repeated work, sequence, and unchosen openings. Evidence can explain a path without turning it into your forecast.
The clearest finding: similar visible conditions repeatedly lead to different outcomes. The differences live in provenance, perseverance, sequence, luck, missing controls, and facts a biography never records.
The model in one minute
Three sources. Two forces. No formula.
What they brought
What capability, drive, or early skill is documented in the person rather than their surroundings? Assessed on 3,578 paths and never summed with the other layers.
What they were handed
What money, family standing, network, or permission was already in place before the work began? Assessed on 3,578 paths and never summed with the other layers.
What surrounded them
What place, timing, institution, or peer group made the next step available? Assessed on 3,578 paths and never summed with the other layers.
Perseverance changes what happens next.
Repeated work, retries, recovery, and sustained practice are shown only when sources document them. Silence is not scored.
Luck acts across every arrow.
Era, encounters, shocks, gatekeepers, and outcome variance can redirect similar visible paths. They stay visible and unscored.
Therefore: compare mechanisms for information, never identities for a verdict. The same layers do not arrive in the same order, receive the same repetition, or meet the same luck.
Beyond the 0.001%-style stories
The archive already reaches toward broader professional distinction.
1,671 paths—46.7% of the archive—are domain-recognized or specialist-known rather than global icons. That is the first expansion toward the kind of success people informally call the top 0.1%. It is not presented as a population percentile without a field denominator.
Open the legacy 22-field annotation detail
Deeper research detail
A head start and a multiplying capability are not the same thing
The first score describes documented position near the beginning. The second describes capacity later present in the path. One may help produce the other, but neither total tells us exactly how that conversion happened.
Before substantial personal proof
Starting advantages
0–24across 12 conditions
Access or conditions documented near the beginning of the path.
Describes the starting position, not what the person later made of it.Capacity later present in the path
Built or converted leverage
0–25across 10 levers
Multiplying capacity documented later in the path.
Measures what was present, not whether it was inherited, earned, self-built, external, or mixed.Five possible origins of leverageSelf-built — Practice, skill, focus, or distribution accumulated directly.Advantage-enabled — A starting resource made the capability easier to develop.Earned access — Earlier work unlocked institutions, collaborators, capital, or reach.External — Timing, a platform shift, or another structural wave multiplied the work.Mixed — Several origins combined and cannot be cleanly separated.Unresolved — The current evidence cannot distinguish how the capability arose.
Person pages now infer a best-supported origin for every non-zero lever, show the linked evidence, state confidence, and preserve “unresolved” when the biography cannot tell us.
Open the legacy ingredient-frequency atlas
Evidence frequency atlas · 3,578 paths
Which scored ingredients are least often documented?
This measures reviewed-source documentation in a curated notable-outcomes sample—not how common a trait is among successful people or the wider population. Each bar uses the same denominator.
Starting advantage
Inherited audience / network is the least frequently documented scored ingredient.
215of 3,578 profiles
6.0% of this curated sample has clear reviewed-source documentation strong enough for a non-zero score.
- Inherited audience / network215 6.0%
- Family financial platform750 21.0%
- Early online platform808 22.6%
- Parent / family domain841 23.5%
8 additional starting conditions
- Adversity / constraint catalyst1,207 33.7%
- Rare early tools1,250 34.9%
- Prodigy / innate ability1,577 44.1%
- Dedicated mentor / coach1,635 45.7%
- Elite institution pipeline2,351 65.7%
- Exceptional peer / cofounder2,456 68.6%
- Frontier geography3,001 83.9%
- Direct domain exposure3,377 94.4%
- Capital safety1,787 49.9%
- Native distribution1,908 53.3%
- Started serious reps before 202,398 67.0%
- Complementary team2,743 76.7%
6 additional multiplying capabilities
- Elite ecosystem network3,142 87.8%
- Structural wave / timing3,480 97.3%
- Scarce skill depth3,487 97.5%
- Prior reps3,548 99.2%
- Concentration intensity3,557 99.4%
- Domain proximity3,567 99.7%
No ingredient is documented in every reviewed path.
Even domain proximity, the most frequently documented scored signal, has no clear documentation in 11 profiles. Zero means “not clearly documented in reviewed sources,” not proof of absence.
Read the scoring and evidence boundary →Why direct comparison becomes redundant
A matching profile is not a matching path
Comparison is useful for exposing ingredients. It stops being useful when resemblance is mistaken for destiny. Four hidden differences prevent that leap.
The same total can be assembled from entirely different advantages and capabilities.
The same capability may be self-built, advantage-enabled, earned, external, mixed, or unresolved.
Order matters: a collaborator before a product is not equivalent to one met after traction.
Unrepeatable encounters and events can separate paths that look identical in the record.
The useful output is diagnostic, not predictive: what was present, where it may have come from, what is missing, and what can still be built.
Luck is cross-cutting—not residual noise
Luck changes the path without becoming a score
This dataset only contains notable outcomes. It cannot show how many people had similar visible ingredients and did not break through, so assigning a “luck score” would create false precision.
Structural luck
Birthplace, era, family, geography, institutions, and being near the right frontier.
Encounter luck
Meeting a collaborator, mentor, coach, investor, selector, or first customer.
Event luck
An algorithm boost, market shock, competitor failure, injury avoided, or unexpected opening.
Outcome variance
Similar visible inputs can still produce different results for reasons the record cannot recover.
Percentiles need a denominator
How rare is a Bill Gates-level tier in this dataset?
Bill Gates is in T1: a band of 653 global-icon profiles, or the top 18.3% of this selected dataset. The tier does not rank people within that band.
T118.3%
T235.0%
T341.9%
T44.8%
T1 · Global iconranks 1–653top 18.3% tier
T2 · Field-leadingranks 654–1,90718.3–53.3% band from the top
T3 · Domain-recognizedranks 1,908–3,40553.3–95.2% band from the top
T4 · Specialist-knownranks 3,406–3,57895.2–100.0% band from the top
Bell-curve intuition
Most people cluster near the middle.
This mental model expects extremes to be rare and distances between adjacent people to stay fairly modest.
Heavy-tail intuition
A small minority can sit extremely far out.
In many outcome domains, recognition, reach, wealth, citations, or attention are highly unequal. Repeated compounding can turn modest early differences into very large outcome gaps.
This is the power-law lesson—not a fitted power law.
The four tiers compress continuous careers into editorial bands, and this dataset begins after an early breakthrough has already happened. We do not fit a Pareto exponent or claim these four tier counts form a power law. The useful idea is that extraordinary outcomes may live in a long tail where “a little better” in inputs does not imply “a little better” in results.
What the 3.5× ratio cannot tell you
It counts lower-tier paths among 3,578 selected early breakthroughs. It cannot count the much larger, unobserved population who attempted similar work, started later, never had a qualifying milestone, or remained undocumented. So it is not “3.5 people failed for every Bill Gates.” The true population denominator is absent.
Data-backed counterexamples
Similar numbers, visibly different lives
These pairs are selected deterministically from profiles that pass the comparison evidence gate.
10/24 start · 15/25 leverage
Same totals, different observed standing
Identical aggregate scores can conceal different fields, timing, trajectories, and career recognition.
Ankit SinghalFounders / operators · T4What they brought: 2 · What they were handed: 1 · What surrounded them: 2
Co-founded Anthrogen, a YC-backed biotech AI startup, and raised $4.6M in venture capital at age 20.
Roger FedererAthletes · T1What they brought: 2 · What they were handed: 2 · What surrounded them: 3Won his first Wimbledon title in 2003 at age 21, and by age 26 (August 2007) had accumulated 12 Grand Slam singles titles including five consecutive Wimbledon championships, reaching world No. 1 in 2004.
14/25 leverage
Same leverage, different starting position
The same capability total does not reveal how much access preceded it or where each lever came from.
Anthony FuResearchers / independent engineers · T4What they brought: 2 · What they were handed: 0 · What surrounded them: 1
Created VueUse (widely-used Vue composition utilities), Vitest (popular testing framework), and Slidev (presentation framework for developers) by age 26, becoming a core team member of Vue, Nuxt, and Vite.
Malala YousafzaiFounders / operators · T1What they brought: 2 · What they were handed: 2 · What surrounded them: 2Awarded the Nobel Peace Prize in December 2014 at age 17, becoming the youngest-ever Nobel laureate in history, for her advocacy of girls' education under Taliban rule in Pakistan's Swat Valley.
5/24 starting advantage
Same starting total, different later capacity
A similar head start does not determine which capabilities are later built, converted, earned, or encountered.
Brooke HuckabyFounders / operators · T4What they brought: 0 · What they were handed: 3 · What surrounded them: 1
Elected mayor of Arabi, Georgia at age 20 in 2023, becoming the youngest elected female mayor in US history.
Matt CoralloResearchers / independent engineers · T3What they brought: 2 · What they were handed: 0 · What surrounded them: 2Began contributing to Bitcoin Core in early 2011 while still in high school (~age 18), becoming one of the earliest active post-Satoshi Core developers; implemented wallet private-key encryption merged mid-2011 (PRs #232/#352).
Open the legacy four-stage research model
A four-stage path
How to read every story in this project
01
Starting advantages
Access, family context, institutions, geography, mentors, peers, tools, ability, and constraints shape the first available moves.
02
Built or converted leverage
Reps, scarce skill, distribution, teams, timing, focus, runway, and domain proximity create multiplying capacity. Its origin may be built, enabled, earned, external, or mixed.
03
Compounding trajectory
Repeated work, feedback, relationships, and well-timed decisions accumulate into a path that becomes difficult to copy quickly.
04
Observed career standing
The tier summarizes documented career recognition through the data cutoff. It is editorial, not calculated from advantage scores or a forecast.
The outcome ladder
What the tiers actually mean
The early milestone determines who enters this dataset; the tier separately summarizes documented career recognition through the data cutoff. It is an editorial band, not a calculation from advantage or leverage scores.
T1653 people · 18.3%
Global icon
Legendary or globally iconic career standingThe documented career became a durable global reference point, shaped a field, or reached iconic recognition well beyond its immediate domain.
T21,254 people · 35.0%
Field-leading
Dominant figure at the top of a fieldThe documented career reached the top level of its field through major prizes, championships, commercial impact, or sustained elite recognition.
T31,498 people · 41.9%
Domain-recognized
Notable and widely recognized within the domainThe documented career established substantial credibility and recognition among people who follow the field.
T4173 people · 4.8%
Specialist-known
Notable, but primarily known within a nicheThe documented career is notable and the early milestone is unusual, but recognition remains narrower or concentrated among specialists.
Open the legacy score-overlap analysis
What the dataset observes
A real gradient—and no clean dividing line
Averages rise as outcome significance rises. The interquartile ranges show the middle half of each tier, making the overlap visible rather than hiding it behind one number.
Exact score cells · bubble area scales with peopleExact score cells · color = tiers present
Different outcomes repeatedly occupy the same score coordinates.
134 / 180occupied score cells contain profiles from at least two tiers.47 score cells contain all four tiers.
T1 Global iconT2 Field-leadingT3 Domain-recognizedT4 Specialist-known
Bubble area11040
Each square is one occupied score cell. Brighter cells contain more outcome tiers.
Preparing the overlap map…
Starting advantage ↔ better tier correlation
0.41
Positive, but far from deterministic
Built/converted leverage ↔ better tier correlation
0.40
Positive, with substantial unexplained variation
| Outcome tier | People | Avg starting advantage | Middle 50% | Avg built/converted leverage | Middle 50% |
|---|---|---|---|---|---|
| T1 · Global icon | 653 | 8.7 / 24 | 6–11 | 13.6 / 25 | 12–16 |
| T2 · Field-leading | 1,254 | 7.8 / 24 | 6–10 | 12.4 / 25 | 11–14 |
| T3 · Domain-recognized | 1,498 | 5.3 / 24 | 3–7 | 11.0 / 25 | 10–12 |
| T4 · Specialist-known | 173 | 5.9 / 24 | 4–8 | 10.4 / 25 | 9–12 |
Why the overlap matters.T1 advantage totals span 2–19; T4 spans 1–14. T1 leverage spans 6–24; T3 spans 0–19. A high score is positioning, not proof. A low score is friction, not a verdict.
Where the observed profiles differ most
The largest T1–T4 score gaps
These are descriptive differences in this successful-only sample. They are useful places to investigate—not estimates of what caused the outcome.
| Dimension | Layer | T1 average | T4 average | Observed gap |
|---|---|---|---|---|
| Dedicated mentor / coach | Starting advantage | 0.94 / 2 | 0.37 / 2 | +0.57 |
| Scarce skill depth | Built / converted leverage | 1.91 / 3 | 1.14 / 3 | +0.77 |
| Elite institution pipeline | Starting advantage | 1.30 / 2 | 0.81 / 2 | +0.49 |
| Prodigy / innate ability | Starting advantage | 0.85 / 2 | 0.38 / 2 | +0.48 |
| Elite ecosystem network | Built / converted leverage | 1.69 / 3 | 1.08 / 3 | +0.61 |
| Capital safety | Built / converted leverage | 0.62 / 2 | 0.23 / 2 | +0.38 |
| Frontier geography | Starting advantage | 1.06 / 2 | 0.69 / 2 | +0.36 |
| Parent / family domain | Starting advantage | 0.57 / 2 | 0.21 / 2 | +0.36 |
What this supports
- Seeing which conditions repeatedly accompany early breakthroughs.
- Separating what a person brought, what they were handed, and what surrounded them.
- Seeing how perseverance, sequence, and luck redirect superficially similar paths.
- Finding mechanisms worth investigating without turning a person into a template.
What this cannot establish
- That a score caused a person’s success.
- That someone with the same profile will reach the same outcome.
- How often similarly advantaged people failed—the dataset has no control group.
- Whether a person’s undocumented advantage was genuinely absent.
Check your interpretation
Six questions the project must answer clearly
If these answers are not obvious, the product has failed its clarity contract.
Why is direct comparison incomplete?
Totals hide composition, provenance, sequence, and luck. Even a close person-specific match is only surface resemblance—not the probability of reproducing someone’s outcome.
Can I test a specific comparison?
Yes. Every published profile can be read through the same three sources, documented perseverance, sequence, and luck. The exercise dismantles the identity comparison instead of producing a resemblance score.
Where does luck fit?
Across every stage. Structural luck, encounters, events, and outcome variance can redirect the path. Luck remains explicit but unscored because the dataset has no failed control group and cannot recover counterfactuals.
Where did the visible advantages come from?
Every profile starts with the same three layers: what the person brought, what they were handed, and what surrounded them. Detailed capability provenance remains available as secondary research evidence.
Why do superficially similar profiles diverge?
They can share totals but differ in which ingredients compose them, how those ingredients arose, the order of events, and unobserved luck. The counterexample pairs above make that divergence concrete.
What percentile is this person—and how many people were far from them?
Each person page shows a qualitative reach band inside this 3,578-path archive. It does not claim a population percentile or estimate how many similar people failed, because the relevant field denominator and failed control group are absent.
Leave with the argument intact
Use comparison for information, not identity.
Inspect another path if it teaches you something. Stop when it becomes a verdict about your pace or worth.
Browse broader professional paths →