Top1000 ScholarThe method in one view
Scholars are compared within the same category and time window
Top1000 Scholar does not assign one universal rank to a scholar. It calculates separate results for each category and time window in which the scholar has sufficient publication evidence. Raw indicators are converted to within-category percentiles and combined into a 0-100 Impact Score. Exact World and National ranks are published only after stronger evidence and entity-quality gates are satisfied.
Publication year, discipline and document type are considered before impact is aggregated.
No single citation count determines the final result; six complementary components are used.
An exact rank requires stronger category evidence than a statistical percentile alone.
Observed scholarly record
Data scope
The 2026 Edition evaluates more than 388 million source work records together with their authorships, citation links, Topics and affiliation observations. After publication-year, work-type, validity and Topic-evidence filters, more than 188 million distinct works contributed to more than 442 million admitted work-Topic relationships. These source and assignment volumes are not the number of works admitted to every ranking; each category and time window applies its own eligibility rules.
Scored publication cohorts include research articles and reviews from complete publication years. Retracted works, paratext records and invalid or incomplete publication years are excluded from scored cohorts. The incomplete 2026 publication year is not used as a publication cohort.
Three ranking levels
Taxonomy and multi-category works
The public taxonomy contains 20 Major Categories, 305 Subfields and 4,516 Topics. Calculations use stable category identifiers rather than display names alone. Topics with identical or similar names remain distinguishable by identifier and parent classification. Each eligible work contributes to one, two or at most three distinct Topics.
Single-category evidence
If one category is retained for a work, that category receives a membership weight of 1.00.
Multi-category evidence
The strongest distinct category is primary and receives 0.70. Up to two evidence-supported secondary categories share the remaining 0.30.
Category weights are used for normalized impact, detectable non-self citation credit, leadership impact and sustained output. Category D-index uses a simpler rule: once an eligible work is admitted to the selected category, its membership weight is 1 for the D-index calculation.
Two separate calculations
Lifetime and 2021-2025 measure different periods
| Ranking window | Eligible publications | Eligible citations | Interpretation |
|---|---|---|---|
| Lifetime | Eligible works published from 1900 through 2025 | Citations observed through the data cutoff | Career-scale scholarly influence |
| 2021-2025 | Eligible works published during 2021-2025 only | Citations received during 2021-2025 only | Influence of recent output, not recent attention to older work |
The same six component weights are used in both windows, but every raw metric, H-index, D-index and percentile is recalculated from the selected window. Older publications do not contribute to the 2021-2025 Impact Score.
Citation treatment
Detectable self-citations and authorship credit are handled explicitly
Detectable self-citation
A citation is classified as a detectable self-citation when the citing and cited works share at least one identified author. Those citation edges are removed from normalized total impact, detectable non-self citation impact, the H-index component used inside the Impact Score and leadership impact. Citation edges with incomplete author identification cannot always be classified; therefore, "detectable non-self" does not claim that every possible self-citation is known.
Authorship credit
category membership weight / number of eligible authors on the workcategory membership weight × (1 / author position) / sum(1 / all eligible author positions)category membership weight / number of identified first, last or corresponding authorsDuplicate appearances of the same author on one work are collapsed before credit is calculated. Harmonic credit emphasizes earlier listed positions while retaining a positive share for each listed author. Fractional credit provides a team-size-sensitive counterbalance. Leadership credit is shared among identified first, last or corresponding authors.
Two H-index definitions serve different purposes
| Metric | Citations | Counting | Role |
|---|---|---|---|
| Score H-index | Detectable non-self citations | Global eligible works in the selected window | 10% component inside the Impact Score |
| Displayed H-index | Total citations, including self-citations | Full counting for every coauthor | Descriptive value shown on leaderboards, profiles and certificates |
| Category D-index | Total citations, including self-citations | Full counting within the selected category | 30% category-specific component inside the Impact Score |
maximum d such that at least d admitted category works have at least d total citations eachFairer comparison
Impact is normalized by field, year and document type
For each eligible work, detectable non-self citations are log-transformed and compared with works from a similar scholarly environment. The preferred reference cell is Subfield by publication year by work type. If that cell is too small, the calculation follows a predefined sequence of broader reference groups.
N(w) = ln(1 + detectable non-self citations of work w) / mean[ln(1 + detectable non-self citations)] in the selected reference cell| Fallback order | Reference cell | Minimum effective work weight |
|---|---|---|
| 1 | Subfield × year × work type | 200 |
| 2 | Subfield × year | 500 |
| 3 | Major Category × year × work type | 1,000 |
| 4 | Major Category × year | 2,000 |
| 5 | Global × year × work type | 5,000 |
| 6 | Global × year | Final available fallback |
A normalized value near 1 indicates impact near the average log-scaled level of the reference cell. The logarithmic transform reduces the ability of one extremely cited work to dominate an entire scholar score. Topic rankings additionally use Topic-specific reference cells where sufficient evidence is available.
Six-component model
Impact Score combines influence, continuity and contribution
Every raw component is converted to a percentile from 0 to 1 within the same category and ranking window. Higher percentiles are better. The weighted total is multiplied by 100 to produce the Impact Score.
| Component | Weight | Operational definition |
|---|---|---|
| Normalized total impact | 30% | 70% harmonic percentile plus 30% fractional percentile of the summed field-year-type normalized detectable non-self impact. |
| Detectable non-self citation impact | 10% | 70% harmonic percentile plus 30% fractional percentile of detectable non-self citation credit. |
| Global detectable non-self H-index | 10% | Largest h for which at least h eligible works have at least h detectable non-self citations in the selected window. |
| Category D-index | 30% | Ordinary H-index inside the selected Major Category, Subfield or Topic, using total citations and full counting for every coauthor. |
| Leadership impact | 10% | Percentile of normalized detectable non-self impact credited to identified first, last or corresponding authors. |
| Sustained output volume | 10% | Fractional equivalent publication volume converted to a category percentile and rescaled to reach its maximum at the category's 95th percentile. |
100 × (0.30Pnormalized + 0.10Pnonself + 0.10PH + 0.30PD + 0.10Pleadership + 0.10Pvolume)Worked example. If a scholar's component percentiles are 0.90, 0.80, 0.75, 0.85, 0.70 and 0.60, the Impact Score is 100 × (0.27 + 0.08 + 0.075 + 0.255 + 0.07 + 0.06) = 81.0.
Total citations and Total Publications are displayed as contextual classic metrics. They are not additional score components.
Evidence before exact rank
Statistical percentiles and exact public ranks use different gates
The statistical cohort is intentionally broader than the exact public-rank cohort. This preserves a meaningful comparison population while preventing a weak one-paper association from creating an exact category rank.
| Scope and window | Statistical scoring cohort | Minimum category evidence for exact public rank |
|---|---|---|
| Major Category / Subfield, Lifetime | At least 1.0 fractional equivalent work | At least 2.0 fractional equivalent works and Category D-index at least 2 |
| Major Category / Subfield, 2021-2025 | At least 0.5 fractional equivalent work | At least 1.0 fractional equivalent work and Category D-index at least 1 |
| Topic, Lifetime | At least 3 Topic-linked works and Topic membership weight at least 1.0 | The statistical requirements plus Topic D-index at least 2 |
| Topic, 2021-2025 | At least 3 Topic-linked works and Topic membership weight at least 1.0 within the window | The statistical requirements plus Topic D-index at least 1 |
Confirmed non-person entities and records that fail the public author-entity gate are not assigned an exact public rank. Topic windows are released only when the window has at least 100 clean statistical authors and at least 10 scholars who pass the public evidence gates.
Three public outputs
Statistical percentile, World Rank and National Rank answer different questions
- Statistical percentile
- The published scholar's relative position in the complete clean eligible population for the same category and window. The denominator is not reduced to the public Top1000.
- World Rank
- The exact ordered position among scholars who pass the stronger public category-evidence and entity-quality gates. A maximum of 1,000 World ranks is published per category and window.
- National Rank
- A secondary rank calculated only among same-country scholars already included in that category-window World Top1000. It never admits additional scholars.
100 × statistical rank / clean eligible ranking denominatorImpact Score descending; normalized total impact percentile descending; D-index percentile descending; stable author identifier ascending- The first leaderboard page loads 100 scholars; pagination covers the full published World Top1000.
- Category search is restricted to the same published World Top1000.
- Scores and statistical percentiles are still calculated against the complete clean eligible population before the public Top1000 boundary is applied.
- A scholar profile shows only category-window records in which the scholar has an exact published World Top1000 rank.
- Country is assigned from the latest observed primary institution at the release cutoff. A missing country produces no National Rank.
Traceable public records
Certificates, versions and corrections
Each public scholar receives a stable ScholarlyIndex identifier that is independent of the current display name and institution. Source profile identifiers remain in the internal lineage map, while observed affiliations are stored as dated episodes. A later name correction or institution move therefore updates the scholar record without creating a new public identity.
Each ranking release is versioned. A certificate identifies the scholar, category, ranking window, World Rank, National Rank where available, statistical percentile, Impact Score and release identity. Its QR code opens the corresponding public scholar profile. A certificate verifies a result in one release; it does not guarantee that the same rank will remain unchanged after later corrections or data updates.
Corrections to author identity, works, institutions, category assignments or citation records are reviewed with an audit trail. Published changes are applied through a new version rather than silently rewriting an earlier certificate.
Found a possible problem?
Include the scholar name, category, identifier and supporting evidence so the record can be reviewed efficiently.