# Provisional coaching rules: inside/outside body position

Draft version: `damien-reference-draft-3`, 2026-09-14.

Current app behavior: authored, source-bound photo reviews now take precedence over these generic rules. They include whole-body observations and a visually assessed turn direction. The session's explicit rider assignments replace the identity checklist. James subsequently revised the constant-foot/knee assumption: foot placement and torso/hip position may differ, so the app no longer assumes fixed knee flexion or foot contact. The rules below remain the fallback when an applicable authored review is absent.

Requested assumption: treat Damien's visible position as the reference for the paired comparison. This is a user-selected working assumption, not a record of coach approval. The draft gives concrete directions for matching the reviewed visible shape; it does not establish that equal projected angles produce equal physical posture or better riding.

## Inside and outside

“Inside” is the rider's side toward the turn. “Outside” is the opposite side. This describes turn direction, not which limb is nearest the camera or appears on image-left.

| Confirmed turn | Inside elbow/knee | Outside elbow/knee |
|---|---|---|
| Left | Anatomical left | Anatomical right |
| Right | Anatomical right | Anatomical left |
| Unknown | Do not assign | Do not assign |

Keep anatomical joint IDs in the evidence records; translate them into inside/outside labels for the report. Require a confirmed turn direction for both paired views before using these labels. The present pair-level direction can be used only when both photos depict the same turn direction; otherwise this draft withholds the comparison. Never infer turn direction from a session's CCW name.

## Eligibility and interpretation

- Require both rider identities and feature-specific view comparability to be reviewed. The views must show sufficiently similar cornering moments for the comparison to be useful.
- Missing or obscured limbs produce no coaching rule. Do not ask about comparability until the relevant feature is visible and assessable in both photos.
- Elbow rules use three confirmed/corrected visible joints: shoulder, elbow, wrist. Compute `delta = James's included angle − Damien's included angle` before rounding. A larger included angle means a straighter elbow.
- Retain a **10° elbow attention threshold** as an adjustable draft display filter. This is an author-selected heuristic, not a researched biomechanical tolerance, measurement-error estimate, or coach-approved threshold. At exactly +10° use “more bend”; at exactly −10° use “less bend.” Below the threshold, show measurements without a focus or a claim that posture is correct.
- Knee coaching describes **“rotated toward the turn”** versus **“up close to the tank.”** These are qualitative descriptions of visible placement, not synonyms for knee bend. The hip–knee–ankle included angle does not measure thigh rotation, knee opening, or proximity to the motorcycle.
- Knee rules therefore require separately reviewed visual evidence about the knee/thigh and the relevant tank outline in both photos. Record whether James appears more open away from the tank, closer to the tank, similar, or unassessable relative to Damien. Retain the anatomical side and confirmed turn in that record. Merely reviewing hip/knee/ankle points is insufficient.
- For the inside knee, opening away from the tank is toward the turn. For the outside knee, opening away from the tank is away from the turn; do not invert that meaning when labeling the rule.
- No numeric knee rotation threshold is proposed. Preserve any included knee angle in a clearly labeled measurement detail, but do not use its sign or magnitude to trigger these knee-position rules.
- A visible gap or apparent proximity does not establish actual tank contact, pressure, support, or hidden 3D rotation. If those visual relationships cannot be judged, withhold the rule.

## Draft elbow content

Apply each family to the inside and outside elbow separately. Substitute the role, both reviewed angles, and absolute difference.

| Rule family | Trigger | Observation template | Suggested action | Limitation |
|---|---|---|---|---|
| Elbow: more bend | Elbow delta ≥ +10° | “Your {inside/outside} elbow looks {difference}° straighter than Damien's ({subject}° versus {reference}°).” | “Using Damien as the reference for this pair, explore a little more bend in your {inside/outside} elbow. Compare the shoulder–elbow–wrist shape as a whole while keeping comfortable access to the controls.” | “This view cannot establish muscle tension, bar loading, or whether torso position, bike fit, or perspective explains the difference.” |
| Elbow: less bend | Elbow delta ≤ −10° | “Your {inside/outside} elbow looks {difference}° more bent than Damien's ({subject}° versus {reference}°).” | “Using Damien as the reference, explore a little less bend in your {inside/outside} elbow while keeping it relaxed. Do not lock the arm straight to chase the photographed angle.” | “The angle alone does not establish cramped posture or excessive bar pressure, or identify a specific torso adjustment.” |

## Draft knee content

Compare foot placement on the peg, inside hip position, and thigh orientation together. The earlier approximately constant knee-flexion assumption was superseded by James's observation that the reference's foot and whole-body position may differ. Do not assume equal peg contact or infer the cause of a lower knee from its projected angle.

Interpret the requested knee-position comparison as the thigh opening at the hip: toward the turn for the inside leg, or held up close to the tank, typically for the outside leg. Camera perspective and thigh orientation are possible explanations for an apparent included-angle difference; the 2D angle alone cannot identify which explains it. Do not translate that difference into “bend/straighten your knee,” a measured hip rotation, or a recommendation to move the foot or pelvis. If the assumed setup is absent or cannot reasonably be used for the pair, withhold this comparison instead of forcing it into the rule.

These triggers use the app's **Knee placement · visual review** input. They must never be triggered from the existing included knee angle. Knees explicitly marked obscured in either rider are excluded.

| Rule | Required reviewed observation relative to Damien | Draft coaching content |
|---|---|---|
| Inside knee: toward the turn | James's inside knee is visibly closer to the tank / less open toward the turn | “Your inside knee appears held closer to the tank than Damien's. Using his position as the reference, explore opening your inside knee more toward the turn.” |
| Inside knee: closer to the tank | James's inside knee is visibly more open toward the turn | “Your inside knee appears rotated farther toward the turn than Damien's. Using his position as the reference, explore bringing it a little closer to the tank.” |
| Outside knee: closer to the tank | James's outside knee is visibly farther away from the tank | “Your outside knee appears farther from the tank than Damien's. Using his position as the reference, explore holding it up closer to the tank.” |
| Outside knee: less tucked | James's outside knee is visibly closer to the tank / more tucked | “Your outside knee appears tucked closer to the tank than Damien's. Compare a slightly less tucked position with his reference before deciding whether an adjustment is useful.” |

Append to every knee focus: “Compare how the thigh opens at the hip and where the knee sits relative to the tank, together with pelvis and foot placement. This view does not measure hip rotation in degrees or prove foot contact, tank contact or pressure.”

Damien remains the pair-specific reference. Do not automatically recommend opening every inside knee or tucking every outside knee regardless of the reviewed difference. “Similar” emits no focus. “Unassessable” shows the missing evidence.

## Rule identities and ordering

| Stable ID | Role | Trigger kind |
|---|---|---|
| `draft-01-inside-elbow-more-bend` | Inside elbow | Included-angle delta ≥ +10° |
| `draft-02-inside-elbow-less-bend` | Inside elbow | Included-angle delta ≤ −10° |
| `draft-03-outside-elbow-more-bend` | Outside elbow | Included-angle delta ≥ +10° |
| `draft-04-outside-elbow-less-bend` | Outside elbow | Included-angle delta ≤ −10° |
| `draft-05-inside-knee-toward-turn` | Inside knee | Reviewed visual comparison |
| `draft-06-inside-knee-closer-tank` | Inside knee | Reviewed visual comparison |
| `draft-07-outside-knee-closer-tank` | Outside knee | Reviewed visual comparison |
| `draft-08-outside-knee-less-tucked` | Outside knee | Reviewed visual comparison |

All rules use priority tier 1. Stable IDs determine presentation order: inside elbow, outside elbow, inside knee, outside knee. This is editorial order, not a riding-importance ranking. Emit at most one rule per feature and three bullets overall; keep remaining observations in evidence details.

## Example (invented evidence)

**Inside elbow: explore more bend.** Your reviewed angle is 105° and Damien's is 80°, so your inside elbow appears 25° straighter. Assuming Damien is the reference, explore a little more elbow bend and compare the whole arm shape.

**Outside knee: hold it closer to the tank.** A separate visual review shows your outside knee farther from the tank than Damien's. Explore bringing it closer to his visible position. No conclusion about tank pressure follows.

These are example copy, not assessments of a private photo.

## App integration and provenance

This draft is now enabled for provisional use at James's request. The app labels the section **“Provisional coaching”**, names the actual reference rider as assumed correct, and displays the 10° elbow heuristic. Provisional bullets carry `user-assumed-reference` provenance and this draft version; the enclosing report retains turn direction, source evidence and revision. The coach-reviewed production catalog and its `reviewProvenance` gate remain unchanged.

The provisional evaluator is separate from the coach-reviewed rule model. Single subject and Compare views share the same report. Angle comparability controls are hidden for unavailable measurements. Knee placement is an optional persisted field in the source-bound pair context, so older saved contexts decode without it; changing turn direction clears the visual review. Existing angle measurements cannot substitute for knee annotation.

Test left/right-turn role mapping, unknown-direction suppression, elbow ±10° boundaries, unavailable joints, every qualitative knee state, and proof that changing a knee included angle alone never triggers knee-placement advice. Also test changes on either rider, deterministic ordering/cap, source-bound visual reviews, and explicit provisional labeling. Numerical and UI tests validate implementation only. Subsequent coach feedback can replace the threshold, wording, or reference assumption without rewriting raw evidence or user corrections.

## External grounding and limits

Honda's [Proper Riding Position](https://global.honda/en/motorcycle-aftersales/plusone/202312.html) describes relaxed arms/elbows, comfortable control operation, and posture varying with motorcycle type. That informs the control-access wording; it does not establish these angle thresholds or track-specific target angles.

Yamaha Champions Riding School's [Why Body Position Matters at the Track](https://ridelikeachampion.com/why-body-position-matters-at-the-track/) discusses body position changing through different riding tasks. That supports checking the cornering moment before comparing photos; it does not validate matching Damien's included angles.

The eight rules, reference-matching assumption, 10° elbow filter, and editorial ordering above are this draft's proposals, not findings attributed to either source.