# Guitar player fingering 15 — one fret hand

The current score uses the same four-finger preparation check independently on
each string. Settled contacts on other strings are retained after their gates.
The actual complete performance has eight simultaneously held contact sites,
seven distinct fret numbers even allowing same-fret barres, and up to five
post-gate contacts. Twenty of its 162 gestures exceed the earlier study's
provisional 95 mm grip target; the widest is 153.425 mm.

This is a source/schedule finding, not a listening verdict or measured human
anatomy. Every whole gesture has at least one alternative within the existing
24-fret instrument and that declared grip target. Do not change or rerender the
obsolete 213-note proposal; use the actual current 248-note performance.

A new player plan retains every written pitch, onset, velocity, gate, picked
attack and connected arrival. It minimizes changes to string assignments before
a declared geometric movement heuristic. Prepared grips and overlapping outgoing
and incoming contact positions must fit the same 95 mm design target.

The planned controller first presses the new current note, then prepares its
remaining grip after the preceding string's quiet-release period. Behind-note
supports are released after a gesture's final fretting motion. Its last support
is released only after the existing 24 ms post-gate mute, unless a new note on
that string already takes over. This avoids accumulating independent fret hands
or substituting extra picked notes for difficult connected phrases.

This plan is not full finger identity, arm velocity, reacting-hand or unilateral
collision physics. It needs independent schedule tests and actual mechanical,
pitch/work and same-score musical A/B evidence before publication. Keep all
older files/URLs, the full post-gate take, Ruth and backing unchanged. No samples,
paid review APIs, agent playback or deferred listening-feed architecture.

## Rolling successor and actual controller

The first whole-grip proposal and its independent verification are retained.
`rolling-plan.json` instead prepares the current note and only an upcoming lower
stop for a pull-off. Its complete-score route changes 27 picked groups rather
than the first proposal's 48, retaining all 248 notes, 162 picks and 86 connected
arrivals. All active stops leave the unchanged 162 mm neck pickup's 12 mm aperture
inside the speaking segment. No stop in the earlier chosen whole-grip route was
found beyond that aperture; it is not a demonstrated failed audio control.

An independent support-union verifier checks 402 actions at 806 interval
boundaries and their midpoints: at most two simultaneous support sites and a
94.527 mm longitudinal span. This is a schedule proof, not a motor-speed or
anatomical model.

`hand_model.py` drives those contacts in the actual coupled six-string/bridge
solver. Auxiliary preparation/release does not create a written note, pick or
hammer/pull impulse. The original contact timeline exactly reproduces the old
signal, final states, pick log and energy. A five-note rolling test has at most
two actual contacts, 89.222 mm span, no stale post-quiet supports, and bare-string
work residual 3.847e-15. All five loaded pitch windows are confident within
3.951 cents. Removing the auxiliary contact schedule changes the mechanical
signal at the same new route; this is not merely a reporting constraint.

The next musical comparison uses the preserved post-gate four-second control
as A, copied byte for byte, and newly renders B at the same initial state and
source clock. Five of its 20 written notes change string; string assignment and
contact strategy both change. Notes, onsets, velocities, gates, picking,
electrical circuit, amp-input calibration, car-song amplifier and backing stay
fixed. A single post-amp gain matches playback level. Complete-score mechanical
audio validation and convincing tone are not established by these short tests.

These sources require the surrounding frozen studies and instrument modules;
the published source copies are not a standalone synthesis package.

## Preparation timing diagnosis

The first 20-note rolling render failed the unchanged broad pitch gate at the
21.65 pull-off: YIN selected its second harmonic, roughly an octave high. The
stored mechanical/circuit waveform still has a 522.657 Hz periodicity candidate,
but its second harmonic is 12.751 dB above its first. That candidate is not used
to excuse or bypass the failed gate. The failed render source, pitch rows and
voltage are retained. An attempted before-pick merge of every lower support also
failed the actual hand-span guard; it is not a publishable control.

`prepared_plan.py` instead moves the next lower support into the existing
pre-pick ramp only when the complete outgoing/incoming hand fits. This rule uses
score/contact geometry, not waveform phase or per-note pitch results. It retains
the same 27-change route, promotes 25 preparations and delays six. The independent
prepared-schedule verifier checks 377 actions, 756 boundaries and 1,511 states:
at most three sites and 94.527 mm span. `preparation-review.md` records the separate
reviewer's matching six infeasible early preparations. The reviewer is closed.

Preparing that stop earlier did not clear the frozen first-trough YIN alias;
the failed prepared-source snapshot and its result are retained too. It is not
presented as an acoustic fix. The new `pitch_period.py` measures the shortest
local period within 0.01 of the best normalized difference over the same broad
70–1800 Hz range. It does not accept an expected pitch, fold an octave, move the
measurement window or alter audio. The old YIN result remains in the report.

The independent estimator passes 112 known-frequency harmonic/phase/noise
fixtures (maximum 0.438 cents) and all 40 preserved A/B actual note windows. It
rejects silence and does not assign a missing fundamental to an even-only wave.
The stored rejected waveform measures 522.657 Hz at confidence 0.996906 and
-1.966 cents, versus the old detector's 1047.226 Hz. This is a bounded measurement
correction, not proof of nicer tone or universal pitch-estimator correctness.