Grade 5 tests: the two pass marks
9 October 2026, evening. Four strand tests (three forms each) and the end-of-grade test (two forms), every item written fresh for the test, examined by a non-author, read cold by readers who never saw the keys. All four strand tests and the end-of-grade test are gated. Counts below are recounted from the form files when each test commits.Two cards, both about where the bar sits. Nothing else on these tests needs you: every fault the readers found was fixed as text and read again.
2 questions for you below, one tap each; none ruled yet. Everything you have already ruled is applied, not re-asked.
Skill: course-build → review-page → write-like-james @ 61d4f89
Status: 9 October 2026, evening. Four strand tests (three forms each) and the end-of-grade test (two forms), every item written fresh for the test, examined by a non-author, read cold by readers who never saw the keys. All four strand tests and the end-of-grade test are gated. Counts below are recounted from the form files when each test commits.
Two cards, both about where the bar sits. Nothing else on these tests needs you: every fault the readers found was fixed as text and read again.
Decisions I need from you
1. Where should the pass mark sit on the Grade 5 strand tests: 27 of 30, or 26?
The course says: Pass at 27 of 30, Grade 3's provisional figure, carried into Grade 5 unchanged.
Why it matters: each form carries 8 to 11 hard items (judge-a-claim, read-data, two-step chains), the released exams' pitch. With David's own targets for a child who has mastered the course (92 per cent right on medium items, 84 per cent on hard), a typical Grade 5 form (3 easy, 17 medium, 10 hard) gives that child an expected score of 27.0: the bar exactly, so a course-master passes about one time in two. At 26 the same child passes about three times in four. The same question is open on Grade 3's page; one ruling can settle both grades.
Cost of changing it: one number in each form's routing line and in the architecture; no item changes.
My recommendation: 26 of 30, for Grades 3 and 5 alike, keeping the exam's pitch
2. Where should the pass mark sit on the Grade 5 end-of-grade test?
The course says: 43 of 47 points (two forms of 45 items: 36 four-option, 4 typed numbers, 5 one-sentence written items; one written item scores 2), 11 of the 45 on Grade 3 and 4 atoms, counted.
Why it matters: your target is 90 per cent on our exit test as the proxy for a top grade on STAAR, Florida SSA and AZSCI. Ninety per cent of 47 is 42.3; the first whole score at or above it is 43 (91.5 per cent). Grade 4's exit took 33 of 36 (91.7 per cent) on the same rule.
Cost of changing it: one number; no item changes.
My recommendation: 43 of 47, the first whole score at or above 90 per cent
Applied without asking (your rulings already given, or decided for the student and recorded)
From your 8 October rulings: every test item is fresh (no PP100, quiz, practice or recall item reused or re-dressed; the examiners struck 14 to 35 re-dressed lesson scenes per strand that the word sweep could not see, and replaced every one); the exam's standard, not a lesson check; a quarter of the end-of-grade test off-grade and counted; three strand forms taken A then B then C with teaching between; a topic scoring zero routes the child back to that topic whatever the total.
Decided and recorded: two one-sentence written items per strand form are written and held out of the 30 until the grader check (Becky); typed numbers only where the exam would have the child type one (two or three a form); a space for thousands in print, with a comma accepted as a typed tolerance only; key letters at 20 to 30 per cent each with no run of three; the bundler now prints each item's lesson line and drawing for the readers (the Earth & Space and Physics first bundles printed neither, and all six readers said so).
The forms at a glance (easy / medium / hard; typed numbers; recounted at commit)
| Test | Form A | Form B | Form C | Gate |
|---|---|---|---|---|
| Physics (Topics 1–7) | 3 / 16 / 11; 1 typed | 4 / 16 / 10; 1 typed | 1 / 19 / 10; 1 typed | GATED |
| Matter (Topics 8–12) | 4 / 17 / 9; 2 typed | 4 / 17 / 9; 3 typed | 4 / 17 / 9; 2 typed | GATED |
| Earth & Space (Topics 13–22) | 1 / 19 / 10; 2 typed | 2 / 17 / 11; 2 typed | 1 / 20 / 9; 1 typed | GATED |
| Life (Topics 23–34) | 3 / 17 / 10; 2 typed | 3 / 19 / 8; 2 typed | 2 / 18 / 10; 2 typed | GATED |
| End-of-grade (two forms, 45 items / 47 points) | 11 / 27 / 7; 4 typed, 5 written | 11 / 27 / 7; 4 typed, 5 written | — | GATED |
What the readers found, in one line each
- End-of-grade: every key agreed across four reads; two giveaway pairs, one stimulus arguing against its own key, and a clipped food-web drawing; all fixed and re-read clean.
- Life: every key agreed (96 of 96); two giveaway pairs survived the first fix and were caught by the re-readers; fixed and re-read.
- Physics: every key agreed; a two-defensible circuit item and "chemical energy" printed before the item that asks for it; fixed and re-read clean.
- Earth & Space: 89 of 90; one shadow-graph option no longer matched its graph after the examiner retimed the dip; fixed and confirmed on the rendered graph, re-read clean; one item then rested on a fact no lesson taught, so the stem now states it.
- Matter: every key agreed; three mass items whose readings did not add up as written; fixed and re-read clean.
Where things are
Forms grade5/strand_tests/G5S_<PHY|MAT|EAS|LIF>_form<A|B|C>.py and grade5/grade_test/G5X_form<A|B>.py; records under grade5/strand_tests/records/ (BUILD, EXAMINER, ADJUDICATION per strand) and grade5/grade_test/records/; blind bundles, sequestered keys and every reader's answers under each folder's blind/.
Skill: course-build → review-page → write-like-james @ 7610d41 · built 9 October 2026 by build_tools_rulings_page.py