re: pre-register H4 -- is the animation clock frame-counted or time-integrated?

H3 pinned 2 units per guest FRAME. The port needs units per SECOND, and that
decomposition is only valid if the clock counts frames. If it integrates
elapsed time then units/second is the invariant and units/frame is an
artefact of whatever rate the host managed.

Evidence it is time-integrated is already in hand and is the caveat H3 had
to quote: across empty label 5376 the plate's alpha moved +82 where three
adjacent labels each moved +23. 82/23 = 3.57, not an integer number of
ticks, and a fixed per-frame increment cannot produce a fractional multiple
of itself.

Instrument: the draw logger now stamps each frame boundary with
Clock::QueryGuestTickCount() and guest_tick_frequency() -- the GUEST's
timebase, 50 MHz per emulator.cc:225, scalar 1.0 per clock.cc:37. No host
wall clock enters the calculation.

Discriminator, within one run: regress alpha step against the label's guest
duration. Flat => frame-counted. Proportional => time-integrated, and the
slope is the answer.

Predicted before the run: time-integrated with r>0.9, at 60 units per guest
second, accept 55-65, 120 and 30 excluded. Stated honestly as a CONSISTENCY
prediction -- it is what the previous capture's run average implies -- not a
blind one.

Control: the same regression on ptcopyright's ramp, a different element with
a different declared segment. Two elements must agree under the
time-integrated model and need not otherwise.

docs/re/h4-units-per-second-preregistration.md

@@ -0,0 +1,80 @@
+# H4 — units per SECOND: is the animation clock frame-counted or time-integrated?
+
+**Pre-registration. Written 2026-09-01 BEFORE the capture was taken**, per
+[`../agents/TEMPORAL-VERIFICATION.md`](../agents/TEMPORAL-VERIFICATION.md).
+Committed first so the prediction cannot be edited afterwards.
+
+## The question, and why the last answer did not close it
+
+[`h3-units-per-frame-measured.md`](h3-units-per-frame-measured.md) pinned **2
+declared units per guest frame**. The port needs **units per second**, and
+`units/s = (units/frame) × (frames/s)`. That is only a valid decomposition if the
+game's animation clock counts **frames**. If it integrates **elapsed time**, then
+units/second is the invariant, units/frame is an artefact of whatever rate the
+host managed, and multiplying by an assumed 30 fps is meaningless.
+
+There is already evidence it is time-integrated, and it is the anomaly the last
+page had to quote as a caveat: across the empty label 5376 the plate's alpha
+moved **+82** where three adjacent labels each moved **+23**. `82 / 23 = 3.57` —
+**not an integer number of ticks**. A fixed per-frame increment cannot produce a
+fractional multiple of itself. That is suggestive and it is one observation.
+
+## The instrument, and why it is not a wall clock
+
+The draw logger now stamps every frame boundary with
+`Clock::QueryGuestTickCount()` and `Clock::guest_tick_frequency()` — the timebase
+the **guest** reads, 50 MHz (`emulator.cc:225`), with `guest_time_scalar_ = 1.0`
+(`clock.cc:37`, unchanged by any flag this harness passes). A duration computed
+from those is the duration **the game experienced**, which is the only one its own
+integrator could have used. No host wall clock enters the calculation.
+
+⟨canary-source⟩ for the scalar and the frequency; ⟨capture⟩ for everything
+measured with them.
+
+## The discriminator — within one run, no second capture needed
+
+For every consecutive pair of frame labels on `ptbtn00`'s ramp, compute the
+label's **guest duration** and the **alpha step** across it. Then:
+
+| if the clock is | Δα vs guest duration | what it means |
+|---|---|---|
+| **frame-counted** | **flat** — Δα ≈ 23 whatever the frame took | units/frame is the invariant; the port multiplies by an assumed fps |
+| **time-integrated** | **proportional, through the origin** | units/second is the invariant; **units/frame is not a property of the game at all** |
+
+The slope of that line, if it is one, is the answer: `slope × 22 / 255` =
+**units per guest second**.
+
+Frame durations under this emulator vary by a factor of several — the previous
+capture's labels ran from nominal to 3.5× nominal — so the regressor has real
+dynamic range and this is not a null test.
+
+## The predictions, stated before the run
+
+1. **The clock is time-integrated.** Δα correlates with guest frame duration,
+   r > 0.9.
+2. **The rate is 60 units per guest second.** I accept **55–65**. This is *not*
+   independent: it is what the previous capture's run-average implies (27.2
+   labels/s × 2.0–2.25 units/label ≈ 54–61), so it is a consistency prediction,
+   not a blind one, and I say so. What is new is that this measurement is
+   per-label rather than a run average, in guest time rather than host time, and
+   carries a shape test the average cannot.
+3. **120 units/s and 30 units/s are both excluded** if 1 and 2 hold.
+
+**If prediction 1 fails** — Δα flat against duration — then the clock *is*
+frame-counted, `units/s` genuinely depends on the console's frame rate, and the
++82 anomaly needs another explanation. That outcome is reportable and I will
+report it.
+
+## The control that removes the test's own subject
+
+The plate's ramp is short. The same regression is run **independently** on
+`ptcopyright`'s ramp in the same capture — a different element, a different
+declared segment, a different part of the screen. Two elements must give the
+**same** units/second under the time-integrated model and need not under any
+other. Disagreement between them fails the test regardless of either slope.
+
+## What this still will not answer
+
+The **console's** frame rate. Even a confirmed 60 units/s says nothing about
+whether the retail game ran at 30 or 60 fps; it says the port should not need to
+know, because it should drive its clock in seconds and not in frames.