The final two weeks: review more of what you know, add nothing new at the end, and sleep

A church tower's band has booked a peal: three hours of ringing, from memory, on a fixed date. Two weeks out, they ring the method they already know on more practice nights than usual. Nobody starts a new method in the last week: a new method needs many practice nights before it holds. Three days out they ring one short quarter peal, then practise lighter. The night before, everyone sleeps. (The band and its peal are the scene's own.)

In the tower	In the last two weeks
More practice nights on the known method	A higher target on known material
A new method needs many nights	A new card comes back several times
No new method in the last week	Nothing new from T−7
One short quarter peal three days out	One short mock at T−3
A full night's sleep before the peal	Sleep every night

Breaks down when: the band hears every slip as it happens, while your gaps show only when you test yourself. Wrong inference: “I'll notice what I've forgotten”: test it on the card.

In real terms: raise the target early, stop adding at T−7, do less at T−3, and sleep.

EXH-4550 is a synthetic student preparing for ExamHall's final exam (an invented case), 14 days out. Both plans have the same days.

EXH-4550 revision plan · ExamHall final exam in 14 days (invented) · SM-2 scheduling model
  T−14  core deck retention 0.90, goal 0.95 · Interval Modifier 1.00 -> log 0.95 / log 0.90 = 0.49
        card "third normal form": Good at T−11, gap 6 -> next gap 7 days, due T−4 (at 1.00: 15 days, T+4)
  new cards: 24 on T−13, T−10 · all in chapters below target · after T−7: 0
  T−3  one 30-minute mock on the weakest chapter · half volume T−3 to T−1
  nights under 7 h: 0 of 14
  plan rules broken: none
MR !4850 the same 14 days, crammed
  T−14  core deck modifier left at 1.00 · card "third normal form": Good at T−11, gap 6 -> next gap 15 days, due T+4, after the exam
  new cards: 45 on T−6, T−4, T−1 · 3 of 3 batches in chapters already at target · after T−7: 45
  T−3  no mock · full volume to T−1
  nights under 7 h: 4 of 14 (nights after T−5, T−3, T−2, T−1; the last 5 h)
  plan rules broken: target not raised at T−14, new cards after T−7, new cards in chapters at target, no T−3 mock, short nights

The revision plan sets the core deck's modifier at T−14, so a card reviewed at T−11 comes back at T−4 instead of after the exam. Its new cards go only to two weak chapters, and stop well before T−7.

The crammed plan leaves the modifier at 1.00, adds 45 new cards in the last six days, skips the mock, and sleeps short on four of the last five nights.

The model counts days; it cannot say how either student would score. Seven hours is this example's line, not a finding.