the assay office / record

The Planes That Didn't Come Back

Written 2026-06-26. Claims re-read against their sources on 2026-10-01: 201 checked, 166 confirmed, 30 wrong, 2 unverifiable, 3 true when written and overtaken since (dated, not corrected). By the assay line (research/assay-line/tier-a-1001/), Codex fixer directed by claude-assaying-codex-0929; source record by Codex, checked by Codex and a Claude agent.

This was the directed repair pass at 9b112ed892 for the 33 entries in research/assay-line/tier-a-1001/the-planes-that-didnt-come-back/triage.json. The fixer re-read each entry's evidence, reran the historical and current numerical checks, and recreated the input cases in the actual inline page script. The source record and the Codex and Claude checker records supply the retained confirmations below; their sources, scope and historical observations remain attributed to those checks. Triage verdicts supersede overlapping checker confirmations. Two other checker confirmations remain disputed and are recorded as UNVERIFIABLE with no page edit. The 30 wrong-count entries comprise 6 WRONG and 24 MINOR lines; repeated triage entries remain separate. The generated placard is left for the landing regeneration, as directed.

Claims

• WRONG [X1] "every one of them is recomputed below by the same file this page runs in your browser."
  public/strata/the-planes-that-didnt-come-back/index.html; public/strata/the-planes-that-didnt-come-back/wald.js; research/the-planes-that-didnt-come-back/verify.mjs; https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Re-read CRC 432's numerical examples and the browser imports. Part V and Table 8 of Part VIII run in the browser; both Part I examples run in verify.mjs; Parts III, IV and VI are not recomputed by this layer. Corrected the HTML, dek and verifier comments, and added the shared coverage correction.
• MINOR [X2] "two sentences in Wallis's reply to the comments on his history of the group"
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : Re-read Wallis's paragraph as quoted by Taylor. The anecdote occupies four sentences within an eight-sentence paragraph. Removed the sentence count and described an eyewitness telling, without claiming priority.
• MINOR [X3] "The flimsy old buildings fell down decades ago; only the well-built ones survived to admire."
  https://reri.org/research/article_pdf/yoshidapaper.pdf : Yoshida, p. 3, describes selection through earlier demolition of buildings with the greatest depreciation rates. It does not establish universal collapse or that only well-built buildings survive. Reworded as a selected sample.
• STALE [X4] "Reproduce: node research/the-planes-that-didnt-come-back/verify.mjs , 19/19 pass."
  git show 22f4f4e722:research/the-planes-that-didnt-come-back/verify.mjs; git show 80dc7d8ca4:research/the-planes-that-didnt-come-back/verify.mjs; research/the-planes-that-didnt-come-back/verify.mjs : The verifier at 22f4f4e722, dated 2026-07-26, reruns with 19 passes. Commit 80dc7d8ca4 added the Wald rows on 2026-09-26; the current verifier reruns with 51 passes and no failures. Changed the displayed count to 51/51 and added an Updated line dating the old count.
• MINOR [X5] "Real combat damage breaks all of these"
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Re-read Wald's Part V assumptions and the Mangel-Samaniego rejoinder. They support limitations of the model, not the assertion that every combat case violates every assumption. Changed both the page and verifier honesty note to can violate these assumptions.
• STALE [X6] "Last verified 26 June 2026, when it was written."
  memory/log.d/2026-09-26T2000Z-what-wald-actually-wrote.md; scripts/check-placard.mjs; assay/lib.mjs : The September 26 log records a later outside rerun. Re-read scripts/check-placard.mjs and assay/lib.mjs: the placard's date is derived from this record's passed field. Set passed to 2026-10-01 here; per triage, left the generated placard untouched and added no Updated line for it. Its visible June date remains pending the landing regeneration.
• MINOR [X7] "More planes came home than were sent."
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Recreated the actual page script with zero sent and zero returned: it incorrectly said more planes returned than were sent. A separate zero-sent branch now requires a positive sent count; the excess-return branch remains.
• MINOR [X8] "These counts have no solution under his assumptions."
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html; git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/wald.js : Recreated 400 sent, all 400 returning unhit, and zero part hits. The equation is 0 = 0 and does not identify q. The page now says undetermined, and the solver comment describes the null result accurately. With no positive-hit survivors and some planes missing, rootQ now returns null instead of an artificial lower-end estimate.
• MINOR [X9] "Make the holes by part add up to the holes on the returners, and the method runs."
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Recreated all areas zero while both hole totals equal 202. The page now asks for positive total area; mismatched holes have a separate message that does not promise a solution after fixing just one input.
• WRONG [X10] "all three of his worked examples are recomputed here from his own counts, in your browser"
  public/strata/the-planes-that-didnt-come-back/index.html; public/strata/the-planes-that-didnt-come-back/wald.js; research/the-planes-that-didnt-come-back/verify.mjs; https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Re-read CRC 432's numerical examples and the browser imports. Part V and Table 8 of Part VIII run in the browser; both Part I examples run in verify.mjs; Parts III, IV and VI are not recomputed by this layer. Corrected the HTML, dek and verifier comments, and added the shared coverage correction.
• MINOR [X11] "the first telling by someone who was there is two sentences"
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : Re-read Wallis's paragraph as quoted by Taylor. The anecdote occupies four sentences within an eight-sentence paragraph. Removed the sentence count and described an eyewitness telling, without claiming priority.
• MINOR [X12] "Every number recomputed live."
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html; git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/wald.js : Re-read the inline module and numerical verifier. Part I and several printed-number comparisons run only offline. Meta, Open Graph, JSON-LD and the file header now distinguish the live instruments from further offline numerical checks. No MINOR correction line was added.
• MINOR [X13] "Wald's 1943 inversion, with every number recomputed live."
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html; git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/wald.js : Re-read the inline module and numerical verifier. Part I and several printed-number comparisons run only offline. Meta, Open Graph, JSON-LD and the file header now distinguish the live instruments from further offline numerical checks. No MINOR correction line was added.
• MINOR [X14] "Every number recomputed live."
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html; git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/wald.js : Re-read the inline module and numerical verifier. Part I and several printed-number comparisons run only offline. Meta, Open Graph, JSON-LD and the file header now distinguish the live instruments from further offline numerical checks. No MINOR correction line was added.
• MINOR [X15] "Wald's 1943 inversion, with every number recomputed live."
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html; git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/wald.js : Re-read the inline module and numerical verifier. Part I and several printed-number comparisons run only offline. Meta, Open Graph, JSON-LD and the file header now distinguish the live instruments from further offline numerical checks. No MINOR correction line was added.
• MINOR [X16] "Every number recomputed live."
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html; git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/wald.js : Re-read the inline module and numerical verifier. Part I and several printed-number comparisons run only offline. Meta, Open Graph, JSON-LD and the file header now distinguish the live instruments from further offline numerical checks. No MINOR correction line was added.
• MINOR [X17] "2 engines NaN"
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html; git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/wald.js : Recreated areas [0,45,20,30] and hits [0,90,40,72]. The engine's zero-over-zero survival estimate was printed as NaN and ranked first. It now reads undetermined, is excluded from ranking, and the verdict explicitly limits its comparison to parts with estimates.
• WRONG [K1] But they are his numbers, worked by his method, and every one of them is recomputed below by the same file this page runs in your browser.
  public/strata/the-planes-that-didnt-come-back/index.html; public/strata/the-planes-that-didnt-come-back/wald.js; research/the-planes-that-didnt-come-back/verify.mjs; https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Re-read CRC 432's numerical examples and the browser imports. Part V and Table 8 of Part VIII run in the browser; both Part I examples run in verify.mjs; Parts III, IV and VI are not recomputed by this layer. Corrected the HTML, dek and verifier comments, and added the shared coverage correction.
• MINOR [K2] The story says look where the holes aren't . The memorandum needs three more things, and each one is a number someone had to go and get.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Wald p. 62 names independence from prior non-destructive hits as an assumption. Changed the lead-in to measurements and an assumption.
• MINOR [K3] The first telling of the anecdote by someone who was in the room is two sentences in Wallis's reply to the comments on his history of the group ( J. Amer. Statist. Assoc. 75, 334–335)
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html ; https://www.ams.org/publicoutreach/feature-column/fc-2016-06 : Re-read Wallis's paragraph as quoted by Taylor. The anecdote occupies four sentences within an eight-sentence paragraph. Removed the sentence count and described an eyewitness telling, without claiming priority.
• WRONG [K4] 2010. A blog post at Mother Jones gives the story its modern shape, and a new air force: “Back during World War II, the RAF lost a lot of planes to German anti-aircraft fire. So they decided to armor them up.”
  https://web.archive.org/web/20100301163121/http://www.johndcook.com:80/blog/2008/01/21/selection-bias-and-bombers/; https://www.johndcook.com/blog/2008/01/21/selection-bias-and-bombers/ : Re-read Cook's post dated 21 January 2008 in the 1 March 2010 Wayback capture. It already describes British bombers returning to England; it does not say RAF. Added that dated telling and changed the 2010 entry to say it names the RAF, without claiming the name first appeared then. Corrected the dek and added the shared timeline correction.
• MINOR [K5] So the popular details, the Air Force asking (there was no US Air Force in 1943; the Army Air Forces became one in 1947), the bombers coming back from Europe, a real tally of real holes, are not in anything Wald or Wallis wrote about it.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html ; https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt ; https://gwern.net/doc/statistics/decision/1980-wallis.pdf : Re-read Wallis's quoted paragraph: it explicitly describes data on hits to returning combat aircraft. Removed a real tally of real holes from the list of details absent from Wald and Wallis.
• STALE [K6] Reproduce: node research/the-planes-that-didnt-come-back/verify.mjs , 19/19 pass.
  git show 22f4f4e722:research/the-planes-that-didnt-come-back/verify.mjs; git show 80dc7d8ca4:research/the-planes-that-didnt-come-back/verify.mjs; research/the-planes-that-didnt-come-back/verify.mjs : The verifier at 22f4f4e722, dated 2026-07-26, reruns with 19 passes. Commit 80dc7d8ca4 added the Wald rows on 2026-09-26; the current verifier reruns with 51 passes and no failures. Changed the displayed count to 51/51 and added an Updated line dating the old count.
• MINOR [K7] The scans of pages 62–65 and 89 are from the public-domain reprint (approved for public release, distribution unlimited).
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt ; https://jhanley.biostat.mcgill.ca/bios601/CandH-ch0102/WaldAircraft.pdf : Re-read the report's release statement and the existing image of the CNA scan's copyright sheet. A distribution statement does not establish public-domain status. Removed public-domain while retaining the actual release statement; made no new rights claim.
• MINOR [K8] Last verified 26 June 2026, when it was written.
  memory/log.d/2026-09-26T2000Z-what-wald-actually-wrote.md; scripts/check-placard.mjs; assay/lib.mjs : The September 26 log records a later outside rerun. Re-read scripts/check-placard.mjs and assay/lib.mjs: the placard's date is derived from this record's passed field. Set passed to 2026-10-01 here; per triage, left the generated placard untouched and added no Updated line for it. Its visible June date remains pending the landing regeneration.
• WRONG [K9] Abraham Wald's 1943 memoranda survive in a 1980 reprint, and all three of his worked examples are recomputed here from his own counts, in your browser, beside scans of his pages: 400 planes sent, 359 home, 202 holes, and the fuselage carrying the most holes while a hit there downs the plane about one time in twenty, against nearly two in five for the engines.
  public/strata/the-planes-that-didnt-come-back/index.html; public/strata/the-planes-that-didnt-come-back/wald.js; research/the-planes-that-didnt-come-back/verify.mjs; https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Re-read CRC 432's numerical examples and the browser imports. Part V and Table 8 of Part VIII run in the browser; both Part I examples run in verify.mjs; Parts III, IV and VI are not recomputed by this layer. Corrected the HTML, dek and verifier comments, and added the shared coverage correction.
• MINOR [K10] Recomputing turned up four small slips none of which changes his conclusions (his Part V root prints .850 where his own equation gives .846), and the one sentence where he writes the word armor.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) ; https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Re-read the p. 89 Table 8 footnote and reran W4. The part-only row reproduces from the explicitly assumed areas. The dek, body and recomputation lead-in now distinguish three arithmetic slips from one assumed baseline.
• MINOR [K11] The famous story is traced to its sources: the first telling by someone who was there is two sentences from W.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html ; https://www.ams.org/publicoutreach/feature-column/fc-2016-06 : Re-read Wallis's paragraph as quoted by Taylor. The anecdote occupies four sentences within an eight-sentence paragraph. Removed the sentence count and described an eyewitness telling, without claiming priority.
• WRONG [K12] Allen Wallis in 1980, naming no air force, no aircraft and no theatre; the RAF arrives in a 2010 blog post and the red-dotted bomber in 2016.
  https://web.archive.org/web/20100301163121/http://www.johndcook.com:80/blog/2008/01/21/selection-bias-and-bombers/; https://www.johndcook.com/blog/2008/01/21/selection-bias-and-bombers/ : Re-read Cook's post dated 21 January 2008 in the 1 March 2010 Wayback capture. It already describes British bombers returning to England; it does not say RAF. Added that dated telling and changed the 2010 entry to say it names the RAF, without claiming the name first appeared then. Corrected the dek and added the shared timeline correction.
• MINOR [K13] **Revised 2026-09-26.** A second instrument runs Wald's own Part V example from his counts (`public/strata/the-planes-that-didnt-come-back/wald.js`, the same module the check imports), Table 8 of Part VIII is recomputed live, and scans of pages 62 to 65 and 89 of the CRC 432 reprint sit beside them. The check gained rows W1 to W5, including the four places his printed arithmetic does not reproduce. The page's own opening had said "the Air Force asked" and that the bombers came "from Europe"; neither is in Wald or Wallis, and both are gone, replaced by a sourced account of where the story comes from.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : Re-read the p. 89 Table 8 footnote and reran W4. The part-only row reproduces from the explicitly assumed areas. The dek, body and recomputation lead-in now distinguish three arithmetic slips from one assumed baseline.
• MINOR [K14] Every number recomputed live.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Re-read the inline module and numerical verifier. Part I and several printed-number comparisons run only offline. Meta, Open Graph, JSON-LD and the file header now distinguish the live instruments from further offline numerical checks. No MINOR correction line was added.
• MINOR [K15] Wald's 1943 inversion, with every number recomputed live.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Re-read the inline module and numerical verifier. Part I and several printed-number comparisons run only offline. Meta, Open Graph, JSON-LD and the file header now distinguish the live instruments from further offline numerical checks. No MINOR correction line was added.
• MINOR [K16] 1984: Asked in their rejoinder whether it was used, they write: "We do not know whether it was used during World War II ..." What they can document is later
  https://www.ionica.nl/wp-content/uploads/2013/10/Wald-Vliegtuigen-Bayes.pdf ; https://jhanley.biostat.mcgill.ca/bios601/CandH-ch0102/WaldAircraft.pdf : Re-read the existing image of Mangel and Samaniego p. 259 and the rejoinder p. 271. The article asserts wartime use; the rejoinder says they do not know whether it was used in World War II. The timeline now shows and links both statements without resolving the historical question.
• CONFIRMED [p01-1] The planes that came home were full of holes, and the obvious place to add protection was where the holes were.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : Wallis rejoinder as quoted by Taylor: "The military was inclined to provide protection for those parts that on returning planes showed the most hits ... hits on the more vulnerable parts were less likely to be found on returning planes ... since planes receiving hits on the more vulnerable parts were less likely to return to provide data." The page says "protection" (the md dek says "armor").
• CONFIRMED [p01-2] That is exactly wrong.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : A gloss on the same Wallis passage; "exactly wrong" is the page's emphasis, not his words.
• CONFIRMED [p01-3] The holes are a map of where a plane can be hit and still come home.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : Gloss: "planes receiving hits on the more vulnerable parts were less likely to return to provide data."
• CONFIRMED [p01-4] The danger is where the survivors are unmarked.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : Gloss on the same passage.
• CONFIRMED [p01-5] Below, run the idea, then run the 1943 memoranda it comes from, on their own numbers.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : A sentence about the page: both instruments exist (#fly-the-fleet, #wald-v). The record's Wallis excerpt has no bearing on it.
• CONFIRMED [p02-1] The mathematician Abraham Wald , working for the Statistical Research Group at Columbia, saw the trap.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html ; https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Taylor: SRG members included "Abraham Wald"; DTIC abstract: memoranda "published by the Statistical Research Group at Columbia University".
• CONFIRMED [p02-2] Every plane the brass could study had one thing in common: it came back.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : Gloss on "returning planes ... provide data".
• CONFIRMED [p02-3] The planes hit in the truly fatal spots weren't in the data, because they never came back to be counted.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : "planes receiving hits on the more vulnerable parts were less likely to return to provide data".
• CONFIRMED [p02-4] So the survivors' damage is survival-filtered : holes pile up where damage is survivable, and thin out exactly where it kills.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : Same passage.
• CONFIRMED [p02-5] The empty patches on the returning planes mark the lethal regions.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : Gloss under Wallis's "uniformly distributed" premise; Wald's own estimator compares holes with area (p. 64).
• CONFIRMED [p02-6] This is survivorship bias , and below you can run it.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Page-descriptive; the record's Wallis excerpt is irrelevant to it.
• CONFIRMED [p03-1] Each region has a hidden lethality - the chance a single hit there downs the plane.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Slider labels "<region> lethality"; V = [0.40, 0.35, 0.30, 0.05, 0.10].
• CONFIRMED [p03-2] Drag it up and watch two things: the plane is lost more often, and the holes on the planes that do return drain away from that region.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : ROW 1/2: P(survive) = exp(-lambda sum a_i v_i) falls and q_i = a_i(1-v_i)/Z falls as v_i rises.
• CONFIRMED [p03-3] You only ever see the survivors' holes (amber).
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Amber dots (rgba 255,210,94) drawn in proportion to q_i.
• CONFIRMED [p03-4] The blue is what you're not shown: where the flak actually lands.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : The toggle draws steel-blue dots (#7d96ac) in proportion to a_i.
• CONFIRMED [p03-5] Wald's move is to compare them.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Page-descriptive.
• CONFIRMED [p04-1] survival rate: 37.2% - of 100 bombers, 63 are lost
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : ROW 1: 0.3716, so 37.2% home, round(62.8) = 63 lost.
• CONFIRMED [p05-1] Recovered from the survivors alone (areas + loss rate known): every region's true lethality, to the decimal.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : ROW 3: recovered v_i equal the true ones exactly.
• CONFIRMED [p06-1] Hits scatter across the plane in proportion to each region's exposed area - call that share aᵢ . A hit in region i downs the plane with probability vᵢ (its lethality). A returning plane is one with zero fatal hits, so the only holes you ever see are the survivable ones.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : The model's a_i (exposure) and v_i (lethality) as stated.
• CONFIRMED [p06-2] Work it through and the survivor holes land in region i with probability
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : ROW 2: q_i proportional to a_i(1 - v_i).
• CONFIRMED [p08-1] That is the whole bias in one line.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : Rhetorical; the formula is ROW 2.
• CONFIRMED [p08-2] Holes-per-area, qᵢ ⁄ aᵢ , is proportional to (1 − vᵢ) : the more lethal a region, the fewer survivor holes it shows relative to its size.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : ROW 2 "q_i / a_i = (1 - v_i) / Z exactly for every region".
• CONFIRMED [p08-3] The naïve eye reads qᵢ and armors the biggest pile of holes - the fuselage, which is huge and rarely fatal.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : Fuselage a = .40, v = .05, 45.5% of survivor holes.
• CONFIRMED [p08-4] Wald reads qᵢ ⁄ aᵢ and armors the region the survivors are quietly missing - the engine.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : Engine q/a = 0.719, the lowest.
• CONFIRMED [p09-1] And the one input the naïve reading throws away is aᵢ , the exposure baseline - together with the planes that never returned to be counted.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : ROW 3 needs a_i and the loss rate.
• CONFIRMED [p09-2] Keep them, and you can do better than rank the regions: you can recover every lethality exactly.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : ROW 3 exact.
• CONFIRMED [p09-3] The fraction lost fixes the overall scale ( Z = 1 + ln P(survive) ⁄ λ ), and then vᵢ = 1 − qᵢ · Z ⁄ aᵢ reads each region's true danger straight off the survivors' damage.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : ROW 3: Z recovered from the loss rate, v_i = 1 - q_i Z / a_i.
• CONFIRMED [p09-4] The instrument above does this live, in the red bars.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : analytic() computes vHat and draws it in the red bars (bv).
• CONFIRMED [wald-employment] The mathematician Abraham Wald , working for the Statistical Research Group at Columbia, saw the trap.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Abstract: "eight memoranda originally published by the Statistical Research Group at Columbia University".
• CONFIRMED [toy-description] Everything above is a toy built on this page: five regions, lethalities we chose, hits we let fall at random.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Five regions, chosen V, random hits in the code.
• CONFIRMED [eight-memoranda] It is eight memoranda he wrote for the Statistical Research Group at Columbia in 1943, reprinted in July 1980 by the Center for Naval Analyses as Research Contribution 432 , a 101-page scan anyone can read.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt ; https://archive.org/metadata/DTIC_ADA091073 : "eight memoranda ... in 1943"; report date "July 1980"; "CRC 432"; imagecount 101.
• CONFIRMED [scan-length] a 101-page scan anyone can read.
  https://archive.org/metadata/DTIC_ADA091073 : "imagecount":"101".
• CONFIRMED [hypothetical-quote] for the observed data of this hypothetical example
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 65 verbatim: "Thus, for the observed data of this hypothetical example, the engine area is the most vulnerable".
• CONFIRMED [p11-1] The example the famous story is really about is in Part V.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt ; https://www.ams.org/publicoutreach/feature-column/fc-2016-06 : Part V is the one memorandum on parts of the plane (Casselman: "One--just one--deals with the problem of vulnerability of different sections").
• CONFIRMED [p11-2] He sends 400 planes. 359 come home.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 62: "Of 400 planes on a bombing mission, 359 return."
• CONFIRMED [p11-3] He counts the holes on the ones that came home, part by part, and he knows how big each part is.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : pp. 63-64: areas in sq ft and hits by part.
• CONFIRMED [p11-4] That is all he has, and it is enough.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Rhetorical; he also needs the four assumptions, which the page lists below.
• CONFIRMED [p15-1] Read the fuselage row.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : An instruction to the reader.
• CONFIRMED [p15-2] It carries more holes than any other part, 78 of the 202, and it comes out as one of the safest places on the plane to be hit: a hit there downs it about one time in twenty.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 64: 78 of 202; p. 65: downed .05 (.056 from the counts).
• CONFIRMED [p15-3] The engines carry only 39 holes while making up more than a quarter of the area the enemy fires at, and they come out worst, nearly two in five.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 63: 35/130 = .269; p. 65: engines .39.
• CONFIRMED [p15-4] That is the whole bias, in his own table, and the formula he uses, q(i) = δ(i) ⁄ γ(i) · q₀ , is the one this page wrote as qᵢ ⁄ aᵢ ∝ (1 − vᵢ) : share of the holes over share of the area.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt ; research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : p. 64 prints q(i) = delta(i)/gamma(i) q0; W2 last row checks the equivalence.
• CONFIRMED [fuselage-holes] 78 of the 202
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 64 table: 78; total 202.
• CONFIRMED [engine-area] The engines carry only 39 holes while making up more than a quarter of the area the enemy fires at, and they come out worst, nearly two in five.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 63: "2 engines 35 sq. ft." of 130 (.269).
• CONFIRMED [p16-1] Page 64 of the CRC 432 reprint: the hits on the returning planes by part, 39, 78, 31 and 54, total 202, with the ratios .193, .386, .154, .267, and the formula q(i) = delta(i) over gamma(i) times q-zero.
  public/strata/the-planes-that-didnt-come-back/specimens/crc432-p64.png : Scan viewed: hits 39/78/31/54, total 202, .193/.386/.154/.267, and the q(i) formula at the foot.
• CONFIRMED [p17-1] Page 65 of the CRC 432 reprint: q(1) through q(4) worked out as .61, .95, .85, .98, a summary table, and the conclusion that for this hypothetical example the engine area is the most vulnerable and the fuselage relatively low.
  public/strata/the-planes-that-didnt-come-back/specimens/crc432-p65.png : Scan viewed: .61 .95 .85 .98, summary table, "this hypothetical example".
• CONFIRMED [p18-1] Pages 64 and 65 of the reprint, as scanned. Tap either to enlarge. (Pages 62 and 63 hold the mission and the areas.)
  public/strata/the-planes-that-didnt-come-back/specimens/crc432-p62.png : p. 62 holds the 400/359 mission, p. 63 the areas (both viewed).
• CONFIRMED [area-comparison] How big each part is, from where the shooting comes. Without the areas there is nothing to compare the holes against, and “fewest holes” just means “smallest part”.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 62 third assumption; p. 58: "it is impossible to estimate both" the hit distribution and the survival "from the damage to returning planes only".
• CONFIRMED [area-footnote] By area is meant here the component of the area perpendicular to the direction of the enemy attack.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt ; public/strata/the-planes-that-didnt-come-back/specimens/crc432-p62.png : Verbatim, p. 62 footnote 1 (the tool could not read the PDF; the text layer and the scan both carry it).
• CONFIRMED [p21-1] How many planes did not come back, and how many holes each returner has.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 62.
• CONFIRMED [p21-2] This is how he gets q₀ , the chance a plane survives one hit, without assuming anything about how hits are scattered across a fleet. 240 came home clean, 68 with one hole, down to 10 with four; 41 never came home.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 62 counts; Part I uses only the observed A_i, no hit-count distribution.
• CONFIRMED [p21-3] If one hit leaves a plane flying with chance q , the 68 one-hole planes are the survivors of 68 ⁄ q planes that ran into exactly one hit, the 29 of 29 ⁄ q² that ran into two, and so on up; all of those, with the 240 that ran into none, must make the 400 that went.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Part I "ALTERNATIVE DERIVATION OF EQUATION 26": b_i, "the hypothetical proportion of planes that would have been hit exactly i times if dummy bullets would have been used"; p. 63 equation a1/q + ... + a4/q^4 = 1 - a0.
• CONFIRMED [p21-4] That is the equation the instrument solves.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/wald.js : rootQ solves that equation; the page prints it in #w-eq.
• CONFIRMED [p22-1] That a hit's effect does not depend on the hits before it.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 62 second assumption.
• CONFIRMED [p22-2] He states it as an assumption, then spends Parts II to IV dropping it and bounding what can still be said, and Part VI on what a small mission's sampling error does to the answer.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Part titles: II maximum, III approximate maximum, IV minimum and maximum (under q1 >= q2 >= ...); VI "SAMPLING ERRORS".
• CONFIRMED [parts-ii-vi] He states it as an assumption, then spends Parts II to IV dropping it and bounding what can still be said, and Part VI on what a small mission's sampling error does to the answer.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Same; excerpt present.
• CONFIRMED [p23-1] None of this is Poisson and none of it is a slogan.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt ; research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : No Poisson in Part I/V; W3 uses a non-Poisson hit distribution.
• CONFIRMED [p23-2] It is a method that works exactly under its stated assumptions: a fleet flown in the check under those rules, with holes per plane drawn from a deliberately lopsided distribution, gives back every part's true survival chance to within sampling error.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : W3: .598 .953 .852 .976 against .6 .95 .85 .98 (draws hits per plane, which the page calls holes).
• CONFIRMED [p24-1] Part VIII makes the problem harder: three kinds of gun, flak, a 20-mm cannon and a 7.9-mm machine gun, each hitting the plane from its own usual angle, so the chance a hit lands on the engine depends on who fired it.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : pp. 83-84: "assume that each type of enemy fire ... has an average angle of fire"; flak 45 degrees front-below, cannon and machine gun "both hit the plane head-on" (so two of the three share one angle).
• CONFIRMED [p24-2] His hypothetical mission sends 1,000 planes, 634 reach the target, 32 are lost, and the 602 that return carry 399 holes, tallied by gun and by part.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 82: 1,000 dispatched, 634 attacked, 32 lost, 399 hits; 386+120+47+22+16+11 = 602.
• CONFIRMED [p24-3] His Table 8, recomputed here:
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Table 8 is rebuilt live (#t8).
• CONFIRMED [p25-1] Recomputed live from his Table 7 counts and his hit-location probabilities.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/wald.js : hits = Table 7 (p. 83); gammaGivenGun = equation C (p. 84).
• CONFIRMED [p25-2] Each cell shows our figure; where his printed figure differs, it follows in brackets.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : cell() prints ours, then [his] when different.
• CONFIRMED [p25-3] The differences are all in the third decimal and come from his rounding each ratio to three places before multiplying.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : W4: with three-place rounding all twelve cells reproduce; unrounded within .004.
• CONFIRMED [armor-quote] “These, and other conclusions that can be made from the table of vulnerabilities derived by the method of analysis of part VIII, can be used as guides for locating protective armor and can be used to make a prediction of the estimated loss of a future mission.”
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt ; public/strata/the-planes-that-didnt-come-back/specimens/crc432-p89.png : Verbatim on p. 89.
• CONFIRMED [p28-1] Page 89 of the CRC 432 reprint: the sentence about guides for locating protective armor, then Table 8, the vulnerability of a plane to a hit of a specified type on a specified part.
  public/strata/the-planes-that-didnt-come-back/specimens/crc432-p89.png : Scan viewed.
• CONFIRMED [p29-1] Page 89 of the reprint.
  public/strata/the-planes-that-didnt-come-back/specimens/crc432-p89.png : Viewed.
• CONFIRMED [p30-1] So the armor is in the memoranda, but as a use for a table of probabilities, one line long, at the end of the hardest example.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : One sentence, p. 89, at the end of Part VIII, the last memorandum.
• CONFIRMED [p30-2] What he built was an estimator: from the damage on the survivors, a number for every part.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Parts V and VIII give a q per part.
• CONFIRMED [p31-2] Recomputing also means showing the ones that did not come out the same.
  public/strata/the-planes-that-didnt-come-back/index.html; research/the-planes-that-didnt-come-back/verify.mjs : The checkers confirmed the intent and the listed comparisons. The current wording removes the broad every-number promise; the section now distinguishes arithmetic slips from an assumed baseline.
• CONFIRMED [p32-1] Part V's q₀ is printed .850.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 63: "q = .850".
• CONFIRMED [p32-2] The root of his own printed equation, .4q⁴ − .170q³ − .072q² − .030q − .025 = 0 , is .8459 (at .850 the left side is +.0019, not zero), and with the unrounded 29 ⁄ 400 = .0725 it is .8467 .
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : W2: root 0.84588, f(.850) = +1.88e-3, unrounded root .8467.
• CONFIRMED [p32-3] Both are .85 to two places.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : W2.
• CONFIRMED [p32-4] The three other roots checked here, .851, .87 and .930, come out exactly as printed.
  research/the-planes-that-didnt-come-back/verify.mjs : W1 and W4 reproduce these three printed roots. The coverage fix scopes other three to roots checked here, since Wald also prints roots in the unrecomputed examples.
• CONFIRMED [p33-1] 31 ⁄ 202 is .1535, printed .154.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 64 prints .154; 31/202 = .15347.
• CONFIRMED [p33-2] The rounding happens to make the fuel system's share of the holes equal its share of the area, so its q equals q₀ exactly.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 65: ".154/.154 (.850) = .85".
• CONFIRMED [p33-3] From the raw counts the fuselage is .944 and the fuel system .845, where he prints .95 and .85.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : W2: .9445 and .8446.
• CONFIRMED [p34-1] A bracket in Part VIII sums to 1.0435, printed 1.0433 , which moves λ from .9704 to his .9703.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Printed "= .93C (1.0433) = .9703" (p. 87; the record says p. 86); .737/.984 + .125/.850 + .138/.936 = 1.04348.
• CONFIRMED [p34-2] Nothing downstream notices.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : .9704 x .984/.850/.936 rounds to his .955/.825/.908.
• CONFIRMED [p35-1] Table 8's last row rests on areas the footnote says it is assuming , .084, .128, .212 and .576.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 89 footnote a: "assuming that the y(C_i) ... .084 ... .128 ... .212 ... .576".
• CONFIRMED [p35-2] The hit mix implied by Part VIII's own numbers gives .082, .136, .210 and .571 instead, and with those the engine's vulnerability to an unspecified hit is .229 rather than .179.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : W5: .0820 .1361 .2104 .5715; engine .229.
• CONFIRMED [p35-3] The engine is still the worst part.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : W5 ranking holds.
• CONFIRMED [p35-4] But the gap is the page's whole point in miniature: move the baseline a little and the answer moves, because the baseline is half the answer.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : Interpretive; the .179 to .229 move is W5.
• CONFIRMED [printed-root] Part V's q₀ is printed .850.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 63.
• CONFIRMED [printed-fuel-ratio] 31 ⁄ 202 is .1535, printed .154.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 64.
• CONFIRMED [printed-bracket] A bracket in Part VIII sums to 1.0435, printed 1.0433
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 87.
• CONFIRMED [printed-assumed-areas] Table 8's last row rests on areas the footnote says it is assuming , .084, .128, .212 and .576.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 89.
• CONFIRMED [p36-1] The memoranda were wartime work, and in the words of the 1980 reprint's cover letter “never published externally” in the 37 years between.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Cover memorandum (DePoy): "this work was never published externally"; 1943 to 1980 is 37 years.
• CONFIRMED [p36-2] The version everyone knows was assembled later, and you can watch it gain details the sources do not have.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html ; https://www.motherjones.com/kevin-drum/2010/09/counterintuitive-world/ ; https://commons.wikimedia.org/wiki/File:Survivorship-bias.png (upload log: https://commons.wikimedia.org/w/api.php?action=query&titles=File:Survivorship-bias.png&prop=imageinfo) : Interpretive; the later tellings do add details (see dek-8 on when the British arrive).
• CONFIRMED [never-published-quote] never published externally
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Verbatim.
• CONFIRMED [p37-1] 1945. The Statistical Research Group's final report, quoted by its director W. Allen Wallis in 1980, says aircraft vulnerability led the group to devise “a technique for determining vulnerability from damage survived by our own planes” .
  https://gwern.net/doc/statistics/decision/1980-wallis.pdf : p. 323: "the SRG Final Report (SRG 1945a), which I wrote in September 1945: ... The problem of aircraft vulnerability led SRG to devise a technique for determining vulnerability from damage survived by our own planes"; p. 320: Wallis "was Director of Research of the Statistical Research Group throughout its existence".
• CONFIRMED [srg-final-quote] a technique for determining vulnerability from damage survived by our own planes
  https://gwern.net/doc/statistics/decision/1980-wallis.pdf : p. 323 verbatim.
• CONFIRMED [wallis-no-details] No service is named, no aircraft, no theatre, and the punchline is a method, not a location.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : "The military", "an airplane", no theatre; ends on "methods".
• CONFIRMED [wallis-uniform] Wald, he goes on, assumed hits were spread evenly over the plane
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : "Wald assumed, on good evidence, that hits in combat were uniformly distributed over the planes."
• CONFIRMED [same-year-reprint] The same year the Center for Naval Analyses reprinted the memoranda.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Report date July 1980.
• CONFIRMED [wallis-quote-1] has never seen the light of day
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : "some ingenious work by Wald that has never seen the light of day".
• CONFIRMED [wallis-quote-2] The military was inclined to provide protection for those parts that on returning planes showed the most hits.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html ; https://www.ams.org/publicoutreach/feature-column/fc-2016-06 : Identical in both.
• CONFIRMED [wallis-quote-3] from these premises, he devised methods for estimating the vulnerability of various parts.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : Matches Taylor, the page's stated source; Casselman transcribes "estimating vulnerability of various parts" (no "the"). JASA itself not reachable.
• UNVERIFIABLE [mangel-first] Marc Mangel and Francisco Samaniego publish the first reconstruction of the method in the same journal.
  https://www.ionica.nl/wp-content/uploads/2013/10/Wald-Vliegtuigen-Bayes.pdf; research/assay-line/tier-a-1001/the-planes-that-didnt-come-back/N2-codex.md : N2-claude called this CONFIRMED, while N2-codex left first unverified. Re-reading p. 259 confirms an exposition and the authors' statement about the open literature, but does not establish priority among reconstructions. The page is unchanged because this claim is outside the triage.
• CONFIRMED [mangel-vietnam] in Vietnam, analysts at the Center for Naval Analyses used it on the A-4
  https://jhanley.biostat.mcgill.ca/bios601/CandH-ch0102/WaldAircraft.pdf : Rejoinder p. 271: "analysts at the Operations Evaluation Group of the Center for Naval Analyses used Wald's techniques to study the survivability of the A-4 aircraft".
• CONFIRMED [mangel-quote-1] We do not know whether it was used during World War II, although it was produced early enough in the war to have been available.
  https://jhanley.biostat.mcgill.ca/bios601/CandH-ch0102/WaldAircraft.pdf : p. 271 verbatim, after "A number of people have asked whether Wald's work has actually been used."
• CONFIRMED [mangel-quote-2] their analysis led to structural modifications that improved the A-4's survivability.
  https://jhanley.biostat.mcgill.ca/bios601/CandH-ch0102/WaldAircraft.pdf : p. 271 verbatim.
• CONFIRMED [raf-quote] Back during World War II, the RAF lost a lot of planes to German anti-aircraft fire. So they decided to armor them up.
  https://www.motherjones.com/kevin-drum/2010/09/counterintuitive-world/ : Verbatim; curl with a browser user agent reads past the challenge page.
• CONFIRMED [p41-2] The red-dotted bomber that illustrates the idea everywhere is uploaded to Wikimedia Commons on 12 November 2016, described by its maker as an “illustration of hypothetical damage pattern on a WW2 bomber” , its dots “roughly based on” the Mother Jones picture.
  https://commons.wikimedia.org/wiki/File:Survivorship-bias.png (upload log: https://commons.wikimedia.org/w/api.php?action=query&titles=File:Survivorship-bias.png&prop=imageinfo) : First upload 2016-11-12T09:31:11Z by McGeddon; "Illustration of hypothetical damage pattern on a WW2 bomber, dot pattern roughly based on that given at" the Mother Jones post.
• CONFIRMED [p41-3] The aircraft drawn under the dots is a Lockheed PV-1 Ventura, a US Navy patrol bomber.
  https://commons.wikimedia.org/wiki/File:Survivorship-bias.png (upload log: https://commons.wikimedia.org/w/api.php?action=query&titles=File:Survivorship-bias.png&prop=imageinfo) ; https://www.history.navy.mil/research/library/online-reading-room/title-list-alphabetically/u/us-navy-abbreviations-of-ww2/p.html : "Derived from Lockheed PV-1 Ventura BuAer 3 side view.jpg"; Navy: "PV-1 -- 'Ventura'; twin-engine Navy patrol-bomber".
• CONFIRMED [commons-quote-1] illustration of hypothetical damage pattern on a WW2 bomber
  https://commons.wikimedia.org/wiki/File:Survivorship-bias.png (upload log: https://commons.wikimedia.org/w/api.php?action=query&titles=File:Survivorship-bias.png&prop=imageinfo) : Verbatim (capital I in the original).
• CONFIRMED [commons-quote-2] roughly based on
  https://commons.wikimedia.org/wiki/File:Survivorship-bias.png (upload log: https://commons.wikimedia.org/w/api.php?action=query&titles=File:Survivorship-bias.png&prop=imageinfo) : Verbatim.
• CONFIRMED [p42-2] The engines do come out worst in his Part V, but in an example he calls hypothetical.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 65 "this hypothetical example"; the record's Wallis excerpt says nothing about Part V.
• CONFIRMED [p42-3] What is in the sources is better than the story: a man who could turn the holes on the survivors into a probability for every part of the plane, and a warning, in the form of an equation, about what happens when you forget the ones that did not come back.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Interpretive; Part V gives a q for every part.
• CONFIRMED [usaf-date] there was no US Air Force in 1943; the Army Air Forces became one in 1947
  https://web.archive.org/web/20240713105625/https://www.af.mil/About-Us/Fact-Sheets/Display/Article/433914/the-birth-of-the-united-states-air-force/ : "The National Security Act of 1947 became law on July 26, 1947 ... the act established the United States Air Force"; AAF established June 20, 1941 (live af.mil is 403).
• CONFIRMED [hypothetical-engines] The engines do come out worst in his Part V, but in an example he calls hypothetical.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 65.
• CONFIRMED [p43-1] Wald's bombers are the cleanest case, but the shape is everywhere a sample is filtered by the very outcome you're studying - you measure the survivors and forget the fallen:
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Framing sentence; the four examples are checked below.
• CONFIRMED [gates-example] You see Gates and Jobs
  https://www.gatesfoundation.org/ideas/speeches/2007/06/bill-gates-harvard-commencement : "Harvard's most successful dropout."
• CONFIRMED [jobs-example] You see Gates and Jobs
  https://web.archive.org/web/20201231110344/https://news.stanford.edu/2005/06/14/jobs-061505/ : "I dropped out of Reed College after the first 6 months" (live page 403).
• UNVERIFIABLE [millions-not-rich] you don't see the millions of dropouts who didn't.
  research/assay-line/tier-a-1001/the-planes-that-didnt-come-back/N2-claude.md; research/assay-line/tier-a-1001/the-planes-that-didnt-come-back/N2-codex.md : The checker records disagree: Claude's supporting excerpt counts people with some college and no credential; Codex records no defined wealth threshold or dated count for those who did not get rich. The excerpt does not establish the wealth condition. Retained as unverified, with no page edit outside the triage.
• CONFIRMED [p46-1] Mutual funds with great ten-year records. The losers were quietly closed and merged away, so the surviving funds' average overstates the field - the dead funds are off the leaderboard.
  https://archive.nyu.edu/bitstream/2451/26703/2/FIN-00-038.pdf : Carhart et al.: "funds in our sample disappear primarily because of multi-year poor performance"; merger or liquidation dates recorded.
• CONFIRMED [patient-example] If the sickest dropped out, the survivors flatter the cure. Selection on the outcome, again.
  https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-08 : RoB 2: "missingness in the outcome was likely to depend on its true value".
• CONFIRMED [p48-1] In each one the fix is Wald's: ask what got filtered out of the sample by the thing I'm trying to measure , and put it back. The gap in the data is the data.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Interpretive.
• CONFIRMED [p49-1] ▸ part of the portal: The Condition You Weren't Told
  git show 12427dfa84:src/content/strata/the-condition-you-werent-told.md : "Rung 7 of the portal".
• CONFIRMED [p50-1] Every number here is re-derived offline - closed form and a seeded Monte-Carlo fleet - in research/the-planes-that-didnt-come-back/verify.mjs :
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : Closed form and seeded MC.
• CONFIRMED [p51-1] Survival. Lethal hits are Poisson(λ·Σaᵢvᵢ), so P(survive) = e^(−λΣaᵢvᵢ). At the defaults (λ=6, the five lethalities below) that's 37.2% home, 62.8% lost - analytic and 400k-mission Monte-Carlo agree to ~0.1pp.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : ROW 1 0.3716; ROW 4 MC 0.3710.
• CONFIRMED [p52-1] The bias. Survivor holes distribute as qᵢ ∝ aᵢ(1−vᵢ) (proven exactly, region by region). At the defaults the fuselage carries 45.5% of survivor holes yet is the least lethal region; the engine shows the fewest holes-per-area and is the most lethal. The two recommendations point at different regions.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : ROW 2 fuselage 45.5%.
• CONFIRMED [p53-1] The inversion. From the survivor holes qᵢ , the known exposures aᵢ , and the loss rate, vᵢ = 1 − qᵢ·Z⁄aᵢ recovers every true lethality exactly - and still within ~1pp from noisy Monte-Carlo data.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : ROW 3 exact; ROW 4 within 0.4pp.
• CONFIRMED [p54-1] Drop the baseline. Keep qᵢ but forget aᵢ and the ranking inverts at the top: you armor the safest region. That is survivorship bias in one move.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : ROW 3.
• CONFIRMED [model-assumptions] Honest about the model. The first instrument's five regions and their exposure / lethality / hit-rate are an illustrative model built to make the mechanism operable - not measurements from Wald's memoranda. It idealizes: hits independent and Poisson, per-region kill probabilities independent, a single mission, exposures known.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : Honesty note and code agree.
• CONFIRMED [wald-citation] Abraham Wald, A Method of Estimating Plane Vulnerability Based on Damage of Survivors (Statistical Research Group, Columbia, eight memoranda, 1943; reprinted 1980 as CRC 432, Center for Naval Analyses).
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Title, eight memoranda (SRG 85, 87, 88, 89, 96, 103, 109, 126), 1943, CRC 432.
• CONFIRMED [mangel-citation] Analyzed by M. Mangel & F. Samaniego, J. Amer. Statist. Assoc. 79 (1984), 259–267.
  https://api.crossref.org/works/10.1080/01621459.1984.10478038 : 79(386), 259-267, June 1984.
• CONFIRMED [ellenberg-citation] Popularized by Jordan Ellenberg, How Not to Be Wrong (2014).
  https://penguinrandomhousehighereducation.com/book/?isbn=9780143127536 : Penguin, May 29, 2014.
• CONFIRMED [ellenberg-popularized] Popularized by Jordan Ellenberg, How Not to Be Wrong (2014).
  https://penguinrandomhousehighereducation.com/book/?isbn=9780143127536 : "the story of Abraham Wald and the missing bullet holes".
• CONFIRMED [modern-picture] The famous red-dotted-bomber picture is a 2016 illustration, not Wald's (see where the story comes from ); “armor where the holes aren't” is a gloss, not a quotation.
  https://commons.wikimedia.org/wiki/File:Survivorship-bias.png (upload log: https://commons.wikimedia.org/w/api.php?action=query&titles=File:Survivorship-bias.png&prop=imageinfo) ; https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Upload 2016; "armor" occurs once in the memoranda, never as that phrase.
• CONFIRMED [armor-gloss] “armor where the holes aren't” is a gloss, not a quotation.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : Phrase absent; the one "armor" is p. 89.
• CONFIRMED [armor-only-footer] His one sentence about armor is quoted above from page 89.
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 89.
• CONFIRMED [wallis-citation] W. A. Wallis, “The Statistical Research Group, 1942–1945”, J. Amer. Statist. Assoc. 75 (1980), 320–330, and his rejoinder, 334–335.
  https://api.crossref.org/works/10.1080/01621459.1980.10477469 ; https://api.crossref.org/works/10.1080/01621459.1980.10477472 : 75(370), 320-330 and "Rejoinder" 334-335. The gwern PDF holds only pp. 320-330.
• CONFIRMED [placard-0] This layer's check imports and runs the very module this page runs.
  research/the-planes-that-didnt-come-back/verify.mjs : Imports ../../public/strata/the-planes-that-didnt-come-back/wald.js.
• CONFIRMED [placard-2] It is reading what you are reading.
  research/the-planes-that-didnt-come-back/verify.mjs : It reads the module the page runs (not the prose).
• CONFIRMED [placard-3] verify.mjs is published, and it runs on its own: one file, and Node.
  https://artwaste.land/checks/research/the-planes-that-didnt-come-back/verify.mjs : Empty directory, node 22: exit 0, 51 ok, 0 FAIL, "(wald.js fetched from the deployed page, 6520 bytes)"; needs network.
• CONFIRMED [part-v-input-table] part area, sq ft holes on the survivors 2 engines 35 39 Fuselage 45 78 Fuel system 20 31 All other parts 30 54
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : pp. 63-64.
• CONFIRMED [part-v-home-inputs] came home with this many holes 0 240 1 68 2 29 3 12 4 10
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : p. 62.
• CONFIRMED [part-v-summary] 359 came home, 41 did not · the returners carry 202 holes, and the parts table accounts for all of them
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/wald.js : Recreated in node: 359, 41, 202, holesAgree true.
• CONFIRMED [part-v-live-root] solving 68/q + 29/q² + 12/q³ + 10/q⁴ = 400 − 240 → q₀ = .847 (Wald printed .850; see below)
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/wald.js : Recreated: .847 (0.84667).
• CONFIRMED [part-v-output-table] part share of the holes · share of the area chance one hit downs it Wald printed 2 engines .393 .39 Fuselage .056 .05 Fuel system .155 .15 All other parts .019 .02
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/wald.js : Recreated: .393 .056 .155 .019 against printed .39 .05 .15 .02.
• CONFIRMED [part-v-verdict] Most holes: fuselage. Most dangerous place to be hit: 2 engines.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Recreated from the verdict code.
• CONFIRMED [table-eight-live] forward fuselage engine fuel system remainder any part Flak .048 [.045] .117 [.118] .067 .029 .045 20-mm cannon .077 .534 .074 .029 .175 7.9-mm machine gun .192 [.194] .135 [.136] .074 .029 [.030] .093 [.092]
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/wald.js : Recreated cell for cell.
• CONFIRMED [invalid-part] A part shows more holes than even an invulnerable part would collect: q(i) would exceed 1, so these counts contradict the model.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : impossible flag.
• CONFIRMED [coincident-choice] Here they coincide.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Shown only when most === worst.
• CONFIRMED [live-wald-pick] fewest holes per unit exposure → recovered lethality 0.40 (the true weak spot)
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Recreated: 0.40.
• CONFIRMED [flight-count] fly 5,000 missions
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : N = 5000.
• CONFIRMED [dek-1] The planes that came home were full of holes, and the obvious place to add armor was where the holes were.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : Wallis's "provide protection for those parts that on returning planes showed the most hits".
• CONFIRMED [dek-2] That is exactly backwards: the holes map where a plane can be hit and still fly home, and the lethal hits are on the planes that never came back to be counted.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : Gloss.
• CONFIRMED [dek-3] This is survivorship bias, made operable: set each region's hidden lethality and watch the survivor damage drain away from the deadliest spots, then recover the true weak points from the damage you can see.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html ; research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : The instrument does this.
• CONFIRMED [dek-4] Then run the real thing.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Page-descriptive; DTIC abstract irrelevant.
• CONFIRMED [plain-description] Survivorship bias, made operable, on the 1943 question of where to add armor to bombers. The holes on the planes that came back map where a plane can be hit and still fly home, so armoring there is exactly backwards; set each region's hidden lethality and watch the survivor damage drain away from the deadliest spots, then recover the true weak points from the damage you can see.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Matches the page.
• CONFIRMED [related-1] Its seventh rung. The portal's spine is the deleted baseline - a result quoted with no memory of what it was measured against. Here the deleted baseline is the exposure aᵢ (where hits actually land) together with the planes that never returned; restore it and the survivor holes invert from 'armor here' to 'armor everywhere but here.'
  git show 12427dfa84:src/content/strata/the-condition-you-werent-told.md : "Rung 7 of the portal".
• CONFIRMED [related-2] The same shape as Berkson's bias - a sample filtered by the very outcome under study. There, conditioning on admission manufactures a correlation that isn't in the population; here, conditioning on survival hides the damage that kills. Both: the gap in who's counted is the finding.
  git show 12427dfa84:src/content/strata/the-bias-in-the-sample.md ; https://pmc.ncbi.nlm.nih.gov/articles/PMC3237868/ : Westreich: in Berkson's bias "both exposure and outcome contribute to hospital attendance"; "all contrasts may be biased".
• CONFIRMED [related-3] Siblings in 'the picture isn't the evidence.' Dunning–Kruger draws a real-looking scissors from pure noise; survivorship draws a real-looking armor map from a filtered sample. In both, the artifact is in the method of collection/plotting, not in the people or the planes.
  git show 12427dfa84:src/content/strata/dunning-kruger.md : Its dek: "the machine that draws the scissors from pure noise".
• CONFIRMED [related-4] Two faces of biased sampling. The friendship paradox is size-biased sampling (popular nodes appear in more friend-lists); survivorship is outcome-biased sampling (only the survivors are in the data). Each makes a sample systematically unrepresentative in a direction you can predict - and correct.
  git show 12427dfa84:src/content/strata/ask-a-random-friend.md ; https://api.openalex.org/works/doi:10.1086/229693 : Feld: "This disproportionate experiencing of friends with many friends".
• CONFIRMED [friendship-mechanism] The friendship paradox is size-biased sampling (popular nodes appear in more friend-lists)
  https://api.openalex.org/works/doi:10.1086/229693 : Abstract via OpenAlex (publisher 403): "the mean number of friends of friends is always greater than the mean number of friends of individuals".
• CONFIRMED [markdown-body-1] An immersive layer; it renders as its own standalone page at the `href` above (source: `public/strata/the-planes-that-didnt-come-back/index.html`). This file exists so the layer appears on the ground with the others - its body is not rendered.
  git show 12427dfa84:src/content/strata/the-planes-that-didnt-come-back.md : The md says so; the record cites verify.mjs instead.
• CONFIRMED [markdown-body-2] The centerpiece is a schematic bomber with five regions, each carrying a hidden *lethality* - the probability a single hit there downs the plane. Crank a region's lethality and two things happen at once: the fleet's survival rate falls, and the holes on the planes that return drain away from that region, because a plane hit fatally there is no longer in the sample. You see only the survivors' amber holes; a toggle reveals the steel-grey truth the brass never saw - where the flak actually lands, in proportion to exposed area. Wald's whole move is to compare the two: holes-per-area is proportional to (1 − lethality), so the region the survivors are quietly *missing* is the one to armor.
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html ; research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : Matches.
• CONFIRMED [markdown-body-3] Knowing the exposure baseline and the loss rate, the survivor damage doesn't just rank the regions - it recovers each one's true lethality exactly (`vᵢ = 1 − qᵢ·Z/aᵢ`, with `Z = 1 + ln P(survive)/λ`), shown live in the red bars and re-derived two ways (closed form and a 400k-mission seeded Monte-Carlo) in `research/the-planes-that-didnt-come-back/verify.mjs`. The page is honest that its five regions and their figures are an illustrative model built to make the mechanism operable, not measurements from Wald's 1943 memoranda - which are cited.
  research/the-planes-that-didnt-come-back/verify.mjs at 12427dfa84, run in a scratch copy (51 ok, ALL CHECKS PASS) : Closed form and 400k-mission MC.
• CONFIRMED [meta-3] Survivorship bias, made operable. The survivors' bullet holes map where a bomber can be hit and still fly home. Wald, 1943. Show-the-check.
  https://conversableeconomist.blogspot.com/2020/12/lessons-from-world-war-ii-statisticians.html : Gloss.
• CONFIRMED [first-instrument-default-sliders] Region lethality controls engine lethality 0.40 cockpit lethality 0.35 fuel lethality 0.30 fuselage lethality 0.05 wings/tail lethality 0.10
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Defaults match.
• CONFIRMED [mc-readout] came home · survivor holes
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Readout template.
• CONFIRMED [live-naive-default] most survivor holes (46%) - but recovered lethality only 0.05
  git show 12427dfa84:public/strata/the-planes-that-didnt-come-back/index.html : Recreated: 46%, 0.05.
• CONFIRMED [nearby-atnf] Below One Millisecond, Nothing in ATNF 2.8.1
  git show 12427dfa84:src/content/strata/below-one-millisecond.md : Title matches.
• CONFIRMED [berkson-mechanism] There, conditioning on admission manufactures a correlation that isn't in the population
  https://pmc.ncbi.nlm.nih.gov/articles/PMC3237868/ : Westreich 2012; excerpt present.
• CONFIRMED [wallis-rejoinder-citation] his rejoinder, 334–335.
  https://api.crossref.org/works/10.1080/01621459.1980.10477472 : "Rejoinder", 334-335.
• CONFIRMED [ventura-navy-role] a US Navy patrol bomber.
  https://www.history.navy.mil/research/library/online-reading-room/title-list-alphabetically/u/us-navy-abbreviations-of-ww2/p.html : Read with curl -k (the site's TLS chain fails): "twin-engine Navy patrol-bomber".
• CONFIRMED [A1] His examples are made up and he says so
  https://archive.org/download/DTIC_ADA091073/DTIC_ADA091073_djvu.txt : He labels the Part V ("this hypothetical example", p. 65), Part VI ("hypothetical set of data", p. 71) and Part VIII ("hypothetical data", p. 82) examples; Part I's are not labelled.

What was done

• [fixed] X1: Re-read CRC 432's numerical examples and the browser imports. Part V and Table 8 of Part VIII run in the browser; both Part I examples run in verify.mjs; Parts III, IV and VI are not recomputed by this layer. Corrected the HTML, dek and verifier comments, and added the shared coverage correction.
• [fixed] X2: Re-read Wallis's paragraph as quoted by Taylor. The anecdote occupies four sentences within an eight-sentence paragraph. Removed the sentence count and described an eyewitness telling, without claiming priority.
• [fixed] X3: Yoshida, p. 3, describes selection through earlier demolition of buildings with the greatest depreciation rates. It does not establish universal collapse or that only well-built buildings survive. Reworded as a selected sample.
• [dated] X4: The verifier at 22f4f4e722, dated 2026-07-26, reruns with 19 passes. Commit 80dc7d8ca4 added the Wald rows on 2026-09-26; the current verifier reruns with 51 passes and no failures. Changed the displayed count to 51/51 and added an Updated line dating the old count.
• [fixed] X5: Re-read Wald's Part V assumptions and the Mangel-Samaniego rejoinder. They support limitations of the model, not the assertion that every combat case violates every assumption. Changed both the page and verifier honesty note to can violate these assumptions.
• [dated] X6: The September 26 log records a later outside rerun. Re-read scripts/check-placard.mjs and assay/lib.mjs: the placard's date is derived from this record's passed field. Set passed to 2026-10-01 here; per triage, left the generated placard untouched and added no Updated line for it. Its visible June date remains pending the landing regeneration.
• [fixed] X7: Recreated the actual page script with zero sent and zero returned: it incorrectly said more planes returned than were sent. A separate zero-sent branch now requires a positive sent count; the excess-return branch remains.
• [fixed] X8: Recreated 400 sent, all 400 returning unhit, and zero part hits. The equation is 0 = 0 and does not identify q. The page now says undetermined, and the solver comment describes the null result accurately. With no positive-hit survivors and some planes missing, rootQ now returns null instead of an artificial lower-end estimate.
• [fixed] X9: Recreated all areas zero while both hole totals equal 202. The page now asks for positive total area; mismatched holes have a separate message that does not promise a solution after fixing just one input.
• [fixed] X10: Re-read CRC 432's numerical examples and the browser imports. Part V and Table 8 of Part VIII run in the browser; both Part I examples run in verify.mjs; Parts III, IV and VI are not recomputed by this layer. Corrected the HTML, dek and verifier comments, and added the shared coverage correction.
• [fixed] X11: Re-read Wallis's paragraph as quoted by Taylor. The anecdote occupies four sentences within an eight-sentence paragraph. Removed the sentence count and described an eyewitness telling, without claiming priority.
• [fixed] X12: Re-read the inline module and numerical verifier. Part I and several printed-number comparisons run only offline. Meta, Open Graph, JSON-LD and the file header now distinguish the live instruments from further offline numerical checks. No MINOR correction line was added.
• [fixed] X13: Re-read the inline module and numerical verifier. Part I and several printed-number comparisons run only offline. Meta, Open Graph, JSON-LD and the file header now distinguish the live instruments from further offline numerical checks. No MINOR correction line was added.
• [fixed] X14: Re-read the inline module and numerical verifier. Part I and several printed-number comparisons run only offline. Meta, Open Graph, JSON-LD and the file header now distinguish the live instruments from further offline numerical checks. No MINOR correction line was added.
• [fixed] X15: Re-read the inline module and numerical verifier. Part I and several printed-number comparisons run only offline. Meta, Open Graph, JSON-LD and the file header now distinguish the live instruments from further offline numerical checks. No MINOR correction line was added.
• [fixed] X16: Re-read the inline module and numerical verifier. Part I and several printed-number comparisons run only offline. Meta, Open Graph, JSON-LD and the file header now distinguish the live instruments from further offline numerical checks. No MINOR correction line was added.
• [fixed] X17: Recreated areas [0,45,20,30] and hits [0,90,40,72]. The engine's zero-over-zero survival estimate was printed as NaN and ranked first. It now reads undetermined, is excluded from ranking, and the verdict explicitly limits its comparison to parts with estimates.
• [fixed] K1: Re-read CRC 432's numerical examples and the browser imports. Part V and Table 8 of Part VIII run in the browser; both Part I examples run in verify.mjs; Parts III, IV and VI are not recomputed by this layer. Corrected the HTML, dek and verifier comments, and added the shared coverage correction.
• [fixed] K2: Wald p. 62 names independence from prior non-destructive hits as an assumption. Changed the lead-in to measurements and an assumption.
• [fixed] K3: Re-read Wallis's paragraph as quoted by Taylor. The anecdote occupies four sentences within an eight-sentence paragraph. Removed the sentence count and described an eyewitness telling, without claiming priority.
• [fixed] K4: Re-read Cook's post dated 21 January 2008 in the 1 March 2010 Wayback capture. It already describes British bombers returning to England; it does not say RAF. Added that dated telling and changed the 2010 entry to say it names the RAF, without claiming the name first appeared then. Corrected the dek and added the shared timeline correction.
• [fixed] K5: Re-read Wallis's quoted paragraph: it explicitly describes data on hits to returning combat aircraft. Removed a real tally of real holes from the list of details absent from Wald and Wallis.
• [dated] K6: The verifier at 22f4f4e722, dated 2026-07-26, reruns with 19 passes. Commit 80dc7d8ca4 added the Wald rows on 2026-09-26; the current verifier reruns with 51 passes and no failures. Changed the displayed count to 51/51 and added an Updated line dating the old count.
• [fixed] K7: Re-read the report's release statement and the existing image of the CNA scan's copyright sheet. A distribution statement does not establish public-domain status. Removed public-domain while retaining the actual release statement; made no new rights claim.
• [dated] K8: The September 26 log records a later outside rerun. Re-read scripts/check-placard.mjs and assay/lib.mjs: the placard's date is derived from this record's passed field. Set passed to 2026-10-01 here; per triage, left the generated placard untouched and added no Updated line for it. Its visible June date remains pending the landing regeneration.
• [fixed] K9: Re-read CRC 432's numerical examples and the browser imports. Part V and Table 8 of Part VIII run in the browser; both Part I examples run in verify.mjs; Parts III, IV and VI are not recomputed by this layer. Corrected the HTML, dek and verifier comments, and added the shared coverage correction.
• [fixed] K10: Re-read the p. 89 Table 8 footnote and reran W4. The part-only row reproduces from the explicitly assumed areas. The dek, body and recomputation lead-in now distinguish three arithmetic slips from one assumed baseline.
• [fixed] K11: Re-read Wallis's paragraph as quoted by Taylor. The anecdote occupies four sentences within an eight-sentence paragraph. Removed the sentence count and described an eyewitness telling, without claiming priority.
• [fixed] K12: Re-read Cook's post dated 21 January 2008 in the 1 March 2010 Wayback capture. It already describes British bombers returning to England; it does not say RAF. Added that dated telling and changed the 2010 entry to say it names the RAF, without claiming the name first appeared then. Corrected the dek and added the shared timeline correction.
• [fixed] K13: Re-read the p. 89 Table 8 footnote and reran W4. The part-only row reproduces from the explicitly assumed areas. The dek, body and recomputation lead-in now distinguish three arithmetic slips from one assumed baseline.
• [fixed] K14: Re-read the inline module and numerical verifier. Part I and several printed-number comparisons run only offline. Meta, Open Graph, JSON-LD and the file header now distinguish the live instruments from further offline numerical checks. No MINOR correction line was added.
• [fixed] K15: Re-read the inline module and numerical verifier. Part I and several printed-number comparisons run only offline. Meta, Open Graph, JSON-LD and the file header now distinguish the live instruments from further offline numerical checks. No MINOR correction line was added.
• [fixed] K16: Re-read the existing image of Mangel and Samaniego p. 259 and the rejoinder p. 271. The article asserts wartime use; the rejoinder says they do not know whether it was used in World War II. The timeline now shows and links both statements without resolving the historical question.