Week 11 Tutorial: Post-Date Analytics and The "Second Date" Metric

Separate observed agreement from inferred subtext. Audit the time budget, evaluate a frozen profile and prepare a reproducible release.

20 May 2027

Inputs#

Bring the frozen Match probability model candidate (exact text, ratings and evidence), Date transition model (exact/seeded comparison and separate observation export) and Offline handover plan (versioned costs and confirmations). Use the worked lecture notes and 50-slide deck for the derivations.

Download the sixteen-file workbook before class. It executes the earlier protocol functions, imports Week 4’s supported bio and checks Week 5’s actual library plan. Its default is an explicitly labelled teaching snapshot assembled now, not a recovery of your earlier candidate. Evaluate your own Week 5 freeze first; use the worked example as a reproducibility reference. All inputs are synthetic; solo implementation and review are accepted.

Finish with The dashboard congratulates itself: the interactive quiz. Six new cases include event histories, dashboard counts, an activity log and release checks. Apply the week’s reporting and evaluation rules, then submit all six responses to unlock the score and worked explanations. The quiz is practice and carries no course marks.

Follow the 60-minute lecture selection and independent-work plan. Complete the preparation and open the workbook before Thursday. The four blocks below are the entire 90-minute tutorial; use existing lecture or study working for review and refinement. Remaining pack material belongs to the week’s independent study allocation.

90-minute tutorial#

1. 20 minutes — audit the log. Reproduce 90 + 20 + 5 = 115 minutes, reject overlapping elapsed-time accounting and critique 10/115. Keep this log separate from the earlier itinerary. Trace agreed, declined, pending and unobserved labels; the counterpart’s event exactly at cutoff 1086 is outside the half-open window. Recount the separate dashboard’s 25% among all versus 50% among resolved records. Neither “nice time” nor a simulation draw supplies an agreement.
2. 25 minutes — evaluate frozen work. Freeze exact text, ratings, evidence and versions. Run the primary score and conservative rank against all 99 controls; save that result before inspecting alternatives. The reference (4,4,3,4) scores 93.75, rank 2/100 because one control ties. Explain the failed target with the actual comparator row. Completed lecture arithmetic can become a peer or solo evidence review.
3. 20 minutes — change the conditions. Double feasibility’s weight for all 100 entries, dividing by 20. The reference becomes 90, rank 2; the separate (4,4,1,4) probe changes 81.25/rank 5 → 70/rank 22. Then declare the distinct 90-minute availability case. The library plan requires 100 minutes, and the unchanged bio contradicts the new window: case fidelity fails despite rank 2. Keep the primary result, common-weight sensitivity and candidate-only evidence change separate.
4. 25 minutes — release review and quiz. Use fifteen minutes to review the release prepared before class: eight checking its labelled fictional page, exact frozen text, evaluation link, keyboard use, 1920×1080 and 390×844, and resize; seven recording actual results, reproducing into a second new directory and comparing output bytes. Generate the worked example beforehand with node week-11-worked-examples.mjs --release-dir ./example-rc1, using a new directory. For your own release, retain your actual frozen candidate, inputs and evaluation. There is no contact collection or app integration. Use the final ten minutes for the six-case quiz: about eight minutes to respond and two to review the explanations after all cases are answered.

Checkpoints and debrief#

• Observation: before 1086 the Week 10 branch is pending; its latest completed record is agreed and the meeting ended. A withdrawal remains visible. A declined invitation terminates regardless of assumed utility.
• Evidence: each rating has a quote and reason. A quote-presence check cannot validate the rating; missing venue detail stays missing even when a notebook plan exists.
• Evaluation: report score, number strictly above, ties, rank and denominator. Do not round before ranking or compare new-weight candidates with old-weight controls. Empty resolved denominators are undefined, not zero.
• Changed case: controls contain no source bios, so do not invent re-ratings for their availability. A newly worded invitation needs a new version and review. An impressive rank cannot override a failed constraint.
• Release: the primary failed target stays visible; all sixteen inputs and output files have a retained checksum record. Browser evidence identifies what was actually inspected. A checksum establishes byte identity against that record, not authorship, accessibility or consent.

Deliverable: Follow-up evaluation#

Save the observation labels and cutoff, time-allocation critique, primary and sensitivity results, changed-case failure, and a release candidate with source/build instructions and actual browser evidence. Explain one disappointing number without editing it away. The workbook can be translated into another language or reproduced through a precise manual procedure.

Week 12 starts from this exact candidate to perform clean reproduction, maintenance and software rollback. Keep its source files and record the runtime version. This supports the existing final project’s validation appendix and release record, not another essay or assessment. The final remains 50%, due 28 May 2027 at 5 pm.

Interactive quiz · 6 cases · about 8 minutes + 2 to review

The dashboard congratulates itself

Six evaluation tickets. The dashboard would prefer a different denominator.

Make a decision in every case, then select Reveal answers. You can change any response before submitting. Scores and explanations stay hidden until the whole quiz is answered.

All events, activity logs, ratings and release examples are authored teaching fixtures. The six cases are separate exercises, not records to join into a dataset. Case 4 uses the unchanged 99 Null Island controls; Case 5 explicitly changes availability for a stress test. Alex’s fixed course case remains unchanged. Practice only: no course marks. Responses stay on this page; reloading starts a fresh attempt.

The cases are readable without JavaScript. Enable JavaScript to navigate, submit your responses and unlock the explanations.

Case 01 / 06The cutoff edits the past

Keep reporting windows and latest observations separate

A new completed Week 10 export retains the following events after both arrivals. Its bill is acknowledged. Reconstruct next-date labels using windows [0, 1082) and [0, 1084), with times in minutes since midnight. A teammate wants to replace both historical labels with the latest result.

One authored meeting history · distinct event IDs

Minute	Coded event
1080	Alex agrees
1082	Counterpart agrees
1084	Counterpart withdraws
1088	Alex leaves

Which report preserves the half-open cutoffs and the completed record?

AAt 1082: agreed. At 1084: declined. Include each event at the stated cutoff.BBoth cutoffs are declined because the latest observation should replace every earlier label.CAt 1082: pending. At 1084: agreed. The latest label is declined and the meeting ended; retain the full history.DAt 1082: pending. At 1084: agreed. The latest label must stay agreed because a previous agreement cannot be withdrawn.

Case 02 / 06Eight records leave the denominator

Name the population behind each percentage

A separate dashboard contains 16 distinct fictional meeting records, all coded at 36 hours after their own meeting. A headline writer removes unresolved rows and describes the resulting fraction as agreement among all meetings.

Synthetic dashboard · one common reporting rule

Label	Records
Agreed	5
Declined	3
Pending	6
Unobserved	2

Which report keeps the evidence and denominators intact?

ARecode all eight unresolved rows as declined so the report has no missing observations.BReport 31.25% among all 16 and 62.5% among the eight resolved records, alongside all four counts and the 36-hour cutoff.CReport 62.5% among all meetings; removing unresolved rows does not change the question.DReport only resolved rows, and use 0% whenever a dashboard contains no agreed or declined records.

Case 03 / 06Five benefit units request a percentage

Audit time coverage and convert the rate’s units

A new log records the four non-overlapping activity intervals below. No activity is recorded between minutes 75 and 85. Assign five arbitrary benefit units for this arithmetic exercise. Use the sum of recorded activity durations as the denominator, including both travel legs and composition.

Model supplied
Each recorded duration is end − start. Convert their sum from minutes to hours, then divide five benefit units by those activity-hours. Keep any unallocated gap separate.

What is the benefit rate in units per recorded activity-hour? Enter the number only; this is not a percentage.

Benefit units per recorded activity-hour

Enter one number. A calculator and your lecture notes are welcome.

Case 04 / 06The candidate changes scales mid-race

Reweight every entry and count ties conservatively

A separate frozen arithmetic probe has ratings (2, 3, 4, 3) in clarity/specificity/feasibility/exit order. These are supplied ratings, not a revised biography. Compare primary weights with doubled feasibility using all 99 unchanged Null Island controls. The control counts below were computed on each rule’s own scale.

Controls compared with the probe · candidate makes entry 100

Rule	Strictly above	Tied
Primary	4	6
Double feasibility	1	4

Model supplied
Primary score = 100 × sum(ratings) / 16. Double-feasibility score = 100 × (clarity + specificity + 2×feasibility + exit) / 20. Conservative rank = 1 + above + tied.

Which comparison follows the declared scoring and ranking rules?

APrimary: 75, rank 11/100. Double feasibility: 80, rank 6/100. Rescore all 100 entries and retain both results.BPrimary: 75, rank 5/100. Double feasibility: 80, rank 2/100. Ties do not count against the candidate.CUse the new candidate score of 80 against the original control scores and report rank 5/100.DThe doubled score is 100 because the new numerator should still be divided by 16; label the candidate top 1%.

Case 05 / 06The route fits and the profile still does not

Separate plan feasibility from changed-case fidelity

In a separate availability stress test, the window becomes Friday 5–6:40 pm: 100 minutes. Keep the Week 5 library itinerary of 100 minutes and $14, and the frozen public wording “Meet Friday 5-7 pm: bus-friendly, $20 total.” A teammate argues that exact itinerary fit makes the unchanged profile ready to release for this changed case.

Which decision respects the changed case and the frozen evaluation?

ARelease the unchanged profile because the route exactly fits the new window.BReject the itinerary because a 100-minute plan must be strictly shorter than a 100-minute window.CSilently shorten the frozen wording and invent new availability ratings for all 99 controls, even though their source bios are missing.DThe itinerary passes with zero time slack, but the 5–7 pm claim contradicts the new window. Preserve the original result; any revised text needs a new version and review.

Case 06 / 06The checksum cannot see the page

Distinguish byte identity from a usable release

Two clean builds of a separate fictional release have identical output bytes and matching manifest checksums. The HTML contains the exact frozen candidate paragraph, but a stylesheet sets #candidate-text to display: none. The evaluation link still works. The developer marks “candidate text visible on phone” as passed without opening the page.

Which release review is justified?

AAccept the visibility claim: a matching checksum proves that the frozen text is readable.BByte reproduction passes; visible-text review fails. Fix the presentation, retain the frozen candidate and evaluation, then record actual browser checks and reproduce the revised release.CReplace the frozen candidate with a shorter profile and regenerate its digest under the old version so the original evaluation remains valid.DMark the whole release accessible because its evaluation link works; the candidate paragraph does not need inspection.

Reveal answers

Answer all 6 cases to unlock your score and the worked explanations.