# ROC Pedigree Simulator: audit and design record

Started: 27 October 2026; updated 28 October 2026. Status: independent engine and comparison page implemented; three full-scale runs completed and passed scientific data checks. Final website release checks follow packaging. Existing site baseline: version 60, commit a8baeb4214ba4160761532a458ef8bbfe703feb2. The original simulator remains available for comparison.

## Confirmed requirements

- Prioritize fidelity to Rohde, Olson and Chang (ROC), rather than claiming new or better science.
- Preserve the evangelical biblical framework: Jacob around 2000 BC; Egypt entry 1876 BC; Exodus 1446 BC; biblical family count 70; approximately 600,000 adult Israelite men at the Exodus, with women and children additional.
- Condition on Jacob's lineage surviving to the present. Do not present extinction as a competing historical outcome. Conditioning must be disclosed and must not be implemented by silently reintroducing descendants.
- Publish a separate page named **ROC Pedigree Simulator**, preserving the original page and its engine.
- Include interactive map and timeline comparisons and reproducible scenario metadata.
- User clarification: intermarriage during the Egyptian sojourn must be represented. A stochastic model beginning only at the Exodus is inadequate for that requirement.

## Source basis

1. Rohde, Olson and Chang, Nature 431 (2004), pp. 562–566, DOI https://doi.org/10.1038/nature02842. Local source: `sources/Modelling the recent common ancestry of all living humans Nature Article.pdf`.
2. Supplementary Methods B, especially pp. 1–7: individual lifetimes, mating, geographic hierarchy, migration, population targets and scaling, initialization. Local source: `sources/NatureCommonAncestors-SupplementB (1).pdf`. Author-hosted copy: https://www.stat.yale.edu/~jtc5/papers/CommonAncestors/NatureCommonAncestors-SupplementB.pdf .
3. Supplementary Methods A and the paper's Tables 1–2 concern the simpler graph model, which must be distinguished from the more detailed individual model.
4. Genesis 41:45, 50–52: Joseph, Asenath and their sons; https://www.biblegateway.com/passage/?search=Genesis+41%3A45%2C+50-52&version=KJV .
5. Genesis 46:26–27: the family count excludes sons' wives and includes family already in Egypt; https://www.biblegateway.com/passage/?search=Genesis+46%3A26-27&version=KJV .
6. Exodus 12:37–38: approximately 600,000 men, additional children and a mixed multitude; https://www.biblegateway.com/passage/?search=Exodus+12%3A37-38&version=KJV .

## Main-engine audit

Inspected `sites/jacob-ancestry/app/model.ts`, the main page, its tests, the independent graph benchmark, and the preserved historical-model diagnostic. Findings refer to the production baseline above.

| Area | Current main simulator | ROC reference | Required improvement |
|---|---|---|---|
| Population unit | Twelve regional fractional coverages | Individual people in nested continents, countries and towns (B pp. 1–3, 7) | Explicit people and nesting; report resolution and all reductions |
| Descent | Smooth regional growth equation | Actual parent links; genealogical ancestry inherited from either parent | Child descent equals logical OR of parental descent; no inheritance percentage or DNA interpretation |
| Time | Ten-year coverage steps | Annual births, deaths and overlapping generations (B p. 1) | Annual event ordering, exact calendar convention, no displayed year zero |
| Mortality | Mortality controls modify a growth multiplier | Discrete Gompertz–Makeham mortality, alpha .01, beta 12.5, maximum age 100 (B p. 1) | Apply mortality to individuals; biblical patriarch exceptions documented separately |
| Reproduction | Fertility and lineage-survival multipliers | Women bear adult-surviving offspring at ages 16–40; birth probabilities adjusted for population targets (B pp. 1, 6) | Actual births with eligible parents; explicitly account for ROC's adult-survivor population convention |
| Partner selection | Same-father setting modifies a multiplier | Local father at least mother's age; preference for previously unpaired men; 80% repeat father when available (B pp. 1–2) | Implement father choice and partnership persistence directly |
| Childlessness | Independent sliders reducing lineage spread | Approximately 19% of women and 36% of men are reported outcomes (B p. 2) | Report measured outcomes; do not force them independently of reproduction |
| Migration | Continuous ancestral influx, including fractions far below one person | Discrete individual migration once before adulthood; within-country towns, inverse-square country destinations, dated ports (B pp. 2–5) | Integer migration events, provenance and zero-arrival outcomes; preserve port units |
| Immigrant clustering | Number of towns changes a multiplier | Source-dependent set of five destination towns (B p. 3) | Actual destination-community clustering |
| Population control | Population mainly supplies denominators | Fertility responds to continental targets and country density; child count capped at four in the described balancing mechanism (B pp. 5–7) | Target-driven births; publish realized populations and any need to depart from reference bounds |
| Biblical foundations | Separate ancestry injections at Jacob, Egypt and Exodus; counts multiplied by a survival factor | ROC does not have biblical anchors | Use an explicit Jacob-specific pedigree foundation; preserve count definitions and continuity instead of repeated ancestry injections |
| Egyptian sojourn | No explicit Israelite–Egyptian household process | Not specified by ROC | Separate ancestry and community state; allow intercommunity parentage before Exodus |
| Exodus movement | Adds ancestry shares to Levant and North Africa | Not specified by ROC | Move the Exodus community; retain descendants in other communities and locations; do not move all descendants solely because they descend from Jacob |
| Uncertainty | Deterministic regional calculation with run-level parameter perturbations | Stochastic individual histories | Separate event randomness, assumed historical parameters and computational resolution |
| Universality | Threshold of 99.999% in each regional mean | An ancestor of every living simulated person | Count exactly in each finite simulated population; never equate a rounded mean or simulated result with established universality among actual humans |
| Validation | Separate graph benchmark reproduces Tables 1–2; main tests check bounds and responses | Detailed model additionally has demography, geography and migration mechanisms | Validate the engine actually used by the new page; retain graph benchmark as a separately scoped check |

## Egypt and community membership

The original proposal to initialize the random pedigree at the Exodus has been withdrawn following the user's clarification. The Egyptian sojourn must be included. Known biblical family relationships should be recorded where specified; unrecorded spouses, ages and intermarriage rates must be identified as assumptions.

Maintain separate states for genealogical descent and community affiliation. A child with either parent descended from Jacob is a descendant. Marriage, conversion or adoption of a community identity does not itself change ancestry. Community affiliation affects the scenario's marriage network and Exodus membership, but must not erase descent in households outside that community.

At the Exodus, the biblical count constrains the Israelite community's adult male population. It is not a count of every Jacob-descended person throughout the world. Descendants living outside that community continue where their simulated family history places them. The mixed multitude must not automatically receive a descent flag.

Genesis 46's seventy is a biblical family enumeration, not a complete all-age local census including every spouse. Initializing exactly seventy total people with an arbitrary 50:50 sex ratio would misrepresent that source. A faithful implementation must preserve the enumeration and model additional spouses explicitly, with uncertainty in unknown particulars.

The Egypt growth target requires a feasibility check against the selected reproduction rules. If no parameterization within ROC's reference bounds reaches the biblical population target with the declared initial household and marriage assumptions, expose that mismatch. Do not conceal it with invented people, relabeling, or an unexplained fertility multiplier.

## Computational scale and validation gates

ROC's detailed simulation used 497 countries and 15,059 towns and reached about 55 million simulated people. Supplement B pp. 5–6 explicitly explains why uniformly shrinking populations can distort small populations and rare migration. A small browser simulation is therefore not an interchangeable substitute for the full model.

User-approved approach: benchmark the core first, provide precomputed checked scenarios, and offer live custom calculations only if feasible. The reference-scale runs execute locally; the browser reads saved snapshots. The initial full-scale 530-year benchmark took 1,963 seconds and missed the Exodus acceptance band. Country-local storage then reduced the 100-year initialization benchmark to 42.9 seconds without changing the allocation distribution. Complete runs remain subject to the checks below. No full ROC replication or measured human ancestry claim is made.

Before publishing numerical results:

1. Audit every adopted rule against Supplement B with page references; explicitly list under-specified choices and deviations, including modern extensions beyond ROC's endpoint.
2. Verify maternal/paternal chronology, integer births/deaths/migration, event conservation, exact descent propagation, and lack of ancestry effects on neutral reproduction rules.
3. Independently traverse stored parent links in manageable test pedigrees and compare with forward descent flags.
4. Verify Egyptian mixed-parent households in both directions, descendants remaining in Egypt after the Exodus, and community changes without ancestry changes.
5. Check realized population trajectories, birth-age distributions, mortality survival and offspring distributions. Report failure to match anchors or demographic targets.
6. Use reproducible seeds; retain failed/condition-excluded attempts in technical provenance while presenting results under the user's survival assumption. Never rescue a failed lineage by reseeding it.
7. Check disconnected groups, closed routes, zero migration and rare-port distributions. Avoid migrations through unopened ports and duplicate migration opportunities.
8. Quantify scale and geography sensitivity before interpreting timings. A stable result at one small resolution is insufficient.
9. Keep existing simulator mathematics unchanged; compare assumption differences and computational units explicitly.
10. Test map, timeline, controls, accessible labels, downloads and the published page after the scientific checks.

## Scope of historical evidence

The Jewish population spreadsheet can document locations, dates, community observations and their uncertainty. It does not establish that every recorded community member descended from Jacob, or identify historical intermarriage rates. Conversion and identity records must not be used as automatic ancestry assignments. Biblical assumptions, ROC mechanisms, independently documented settlement evidence and exploratory rate choices must remain distinguishable.

## Independent review and completed-run findings

A separate GPT-6 Astra review inspected the frozen engine and source assumptions before release. It identified no fatal implementation defect requiring the full trials to restart. It requested preservation of source opening dates separately from already-open route sentinels, and a visible explanation of the absence of an additional Jewish endogamy barrier after the Exodus. Both changes were made without changing the frozen engine. This was an AI-assisted code and methods review, not external scholarly peer review or independent replication.

The first two completed 55-million-target runs used seeds 17001 and 17002 and the 1% new-partner crossing assumption. They achieved 573,801 and 575,432 resident community men aged 20+ at Exodus. Adult descendants outside the community numbered 27,535 and 30,044. Both passed the final census, event conservation, chronology, and survival checks without reseeding.

They differ at the endpoint: the first retains 2,947 non-descendant adults in South America; the second has exact adult coverage by its saved AD 1475 snapshot and exact coverage of living modeled juvenile survivors at AD 2026. This is variation between conditional simulated histories, not a measured probability of real human universality. The first saved exact-coverage date can miss an earlier event between snapshots.

The fertility and migration mechanisms do not reproduce every ROC demographic outcome. Female lifetime childlessness is about 16.35% in these trials, versus approximately 19% reported by ROC. Final mainland targets differ by up to about 2.5%; Hawaii and Greenland exceed their small targets by roughly 61–67%, and Easter Island retains a few adults against a zero endpoint target. Actual regional counts and target differences are included in the downloadable diagnostic report. These limitations are preserved rather than hidden by rescaling final results.

The bounded higher-intermarriage continuation stopped at seed 17006 after that run completed and passed the unchanged acceptance rule. It achieved 542,398 Exodus men, 107,052 adult descendants outside the community at Exodus, and 54,896,068 adults at AD 2026. Its first saved exact adult-coverage snapshot is AD 1650; all modeled living juvenile survivors also descend at AD 2026. Earlier 5% attempts 17003, 17004, and 17005 missed the Exodus band and remain excluded in provenance. The 1% group has two accepted runs from two attempts; the 5% group has one from four. These counts are not historical likelihood estimates. No fertility ceiling, count, or acceptance range was changed after inspecting outcomes.

Browser checks with the completed results verified run switching while retaining the chosen date, Exodus and AD 1 milestones, keyboard date movement, 4× playback and automatic endpoint stopping, persistent map selection, and a 390-pixel phone layout. At AD 1 the calculated slider-thumb and date-label centers agreed to under 0.01 pixels. No browser console errors were observed during these checks. Final publication checks and deployment status are recorded separately.