Task: add documented, disprovable estimated birth/death years (rules E1-E5) to ANC entries with an anchor; render faded/dashed estimated bars on the timeline (including ANC-only people not previously drawn); ~ year style in fan/tree/tooltips/table; Estimated-dates block on person pages; methodology section in the tree doc. Static verification only (no browser — reserved).
Candidate sweep result (ANC entries with null b or d + anchor)
22 people get estimates; 4 more ANC-only people with REAL dates get first-time bars (Lumpkin Gaither Foster 1856–1925, Franklin Calloway Foster 1830–1911, Harriett Foster 1826–1880, Mary Barnard 1609–1676 — E4: real dates override, so NO estimates for them even though the dispatch candidate list mentioned Lumpkin/Harriett as estimate candidates; their dates were pinned in the morning research rounds, list was stale).
| ahn | person | eb | ed | rule |
|---|---|---|---|---|
| 23 | Adella Rogers Olney | 1866 | 1928 | E2 (Florence 1889 −23), bound by parents Edgar d.1875 / Mary b.1845 |
| 29 | Mary Adeline Hooper | — (b 1856 real) | 1928 | E3 floored: max(1856+62=1918, 1925+3) |
| 45 | Sarah Mary Salisbury | 1833 | 1895 | E2 (James Preserved 1856 −23); consistent w/ mother d.1840 |
| 49 | Rowena E. Rowland | — (b 1818 real) | 1883 | E3 floored: max(1880, 1880+3) |
| 89 | Lucy Burt | 1794 | 1856 | E1 (m.1815 −21) |
| 101 | Martha Jane Watkins | 1810 | 1872 | E2 (Wm N. 1833 −23) |
| 131 | Mary (Strain) | 1750 | 1812 | E2 (Rachel 1773 −23) |
| 204 | Lemuel Thomasson | — (b 1785 real) | 1852 | E3 floored: max(1847, 1849+3) |
| 353 | Elizabeth Goodspead | 1733 | 1795 | E2 (William 1756 −23) |
| 705 | Phebe Angell | — (b 1697 real) | 1759 | E3 |
| 1408 | William Olney Ⅲ | — (b 1663 real) | 1725 | E3 |
| 1409 | Catharine Sayles | 1671 | 1733 | E1 (m.1692 −21) |
| 1410 | James Angell | 1653 | 1715 | E1 (m.1678 −25) |
| 1411 | Abigail Dexter | 1657 | 1719 | E1 (m.1678 −21) |
| 2817 | Elizabeth March | 1639 | 1701 | E1 (m.1660 −21) |
| 2820 | Thomas Angell | 1626 | 1688 | E2 chained off son James eb 1653 — weakest class |
| 2821 | Alice (surname unproven) | 1630 | 1692 | E2 chained — weakest class |
| 2822 | Gregory Dexter | 1630 | 1692 | E2 chained off daughter eb 1657 — weakest class; 1643 imprint tension flagged |
| 2823 | Abigail Fullerton | 1634 | 1696 | E2 chained — weakest class |
| 5633 | Marie Ashton | — (bapt 1605 real) | 1667 | E3; POSSIBLE pre-1659 record noted as would-be disproof |
| 11266 | James Ashton | 1578 | 1640 | E2 (Marie bapt 1605 −27) |
| 11267 | Alice (Ashton) | 1582 | 1644 | E2 (Marie bapt 1605 −23) |
Decisions / deviations
- Lumpkin, Harriett, Franklin Calloway Foster: NOT estimated — they now carry real documented dates in ANC (E4 real-overrides). They (plus Mary Barnard) get real solid bars via the same FLAT-append path.
- Mary (Strain): dispatch framed her anchor as "husband 1750–1826"; the rules have no husband formula, and she has a legitimate E2 anchor (daughter Rachel b. 1773) → used E2 (eb 1750 — happens to coincide with husband's birth year; basis notes both).
- Martha Jane Watkins: same — used E2 via son Wm N. b. 1833 rather than an un-ruled husband anchor; consistent with husband's 1798–1872 span, noted in basis.
- Thomas Angell/Alice, Gregory Dexter/Abigail Fullerton: estimate-on-estimate (E1 result of child feeds E2 for the parent). Flagged in each basis as the weakest class per dispatch. Gregory Dexter tension flagged: his documented 1643 London imprint of Williams's "Key" implies a birth earlier than the chained ~1630 (a 13-year-old printer is not credible); the basis line says so out loud as a ready disproof. Rules E1-E5 were applied exactly; no ad-hoc formula invented. FLAG FOR PARENT/FOUNDER review.
- Y0 1600 → 1570: James Ashton eb 1578 / Alice eb 1582 precede the old clock start; extended the axis so their bars don't render at negative offsets.
- #canvas min-width 2520 → 3010: 86 → 112 bars at 26px pitch (52 + 112×26 = 2964).
- Timeline alias map added (
Wm N. LeQuire→William N. LeQuire,Mattie Hayes→Mattie Hayes Mashburn): pre-existing latent bug — these GROUPS spellings missed their ANC key, so fOf fell back to 0.99 (far-right column) AND the FLAT-append dedupe would have double-drawn them. Alias fixes genealogical placement + dedupe. - svg (fan) click handler deduplicated into ancTip() — it was a verbatim copy; single point of change for the basis line.
- Story-text touch-ups where text claimed someone is "not drawn"/"no timeline bar" (Thomas Olney re Marie, Wm Russell Olney re Sarah, James Preserved re Adella, Phebe, Rowena "off the clock").
- Footer "Known but not drawn to scale" list: everyone on it is now drawn → replaced with the estimation-methodology sentence. Table's "dates not yet pinned" rows (Adella, Sarah Mary, Marie, Lucy) removed — they're in the main table now with ~ years.
Static verification (results appended at end of run)
node --checkon the extracted inline<script>(77.4KB): PASS (checked after the final index.html edit).- Node data audit (re-eval of ANC/GROUPS via build-people's extractor): 22 estimated entries, all with ebasis, none violating E4 (no eb beside real b / ed beside real d), all ed ≤ 2026, start < end, valid regions; FLAT append = 26 new bars (22 est + 4 real), 86 → 112 total, zero duplicate keys, zero fOf 0.99 fallbacks, earliest start 1578 ≥ Y0 1570, canvas need 2964px ≤ 3010 min-width.
node build-people.mjs: clean — 112 pages + index + slugs.js + style.css, no unassigned-line warning (the six Angell/Dexter people were previously line-less and missing from people/index.html; assigned to Olney & Lambert).- Grep: exactly 22 pages contain the "Estimated dates" block, and they are exactly the estimated set. Zero occurrences of
~null,NaN,~undefined,null–/–nullin index.html or people/*.html. - Samples verified: phebe-angell (1697 – ~1759 + block), lucy-burt (~1794 – ~1856 + block), marie-ashton (1605 – ~1667 + block w/ pre-1659 caveat). Family-link spans on relatives' pages show tilde years (e.g. Adella as (~1866–~1928) on Florence's page).
- Slug map resolves
Name|nullkeys for all sampled estimated people → timeline/fan/tree tooltips will link to their pages. - "dates open" now appears only in confidence badges / documented-facts lines ("Confirmed · dates open") — accurate: real dates ARE still open; the estimate block sits alongside. No stale "not drawn"/"off the clock"/"no timeline bar" copy remains in index.html.
- No browser/Playwright used; no deploy; visual QA left to the parent session.