HARMONY_FIT.md

music/exemplars/instruments/HARMONY_FIT.md

HARMONY FIT -- harmony-fit/1.0

The VERTICAL instrument. It measures whether a melody was written AGAINST a harmony, which is the clause of Josh's round-3 grade -- "they clash ... never any harmonies added on" -- that no other instrument in the battery can see.

What this instrument is, and is not

TIER: score only. HARMONY FIT IS A SCORE INSTRUMENT AND HAS NO AUDIO TIER. Chord-tone agreement, dissonance classification and parallel-perfect detection all require knowing WHICH NOTES ARE SOUNDING, and docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:559-569 puts exactly those on the not-measurable-from-audio list (GAP 1, polyphonic transcription). Pass --plan. A card-only or audio-only tier here would be an invented proxy for the one thing the research says cannot be recovered from a render.
CALIBRATION: NONE. The corpus is audio and there is no score for it, so no axis here is calibrated against the 30 loved tracks. This is a score-side GENERATION-DEFECT detector in the register floor_instruments.structural_yield uses about itself, and its evidence is (1) cited theory, (2) the per-cue random-melody null, (3) the contrast between our own rounds on a mechanism we deliberately changed.
NULL: melodic pitches re-drawn uniformly from the plan's own mode inside each part's own compass, rhythm untouched, 200 seeded draws, seed 20260808; percentile is mid-rank (strictly below plus half the ties)
NOT GATED: chord_tone_share is gated against its own random-melody null and never against a fixed consonance quota: docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:63-64,71-73 put the strict ratio of consonance types on the does-not-survive list

Headline -- does round 3 sit AT its own random-melody null?

round2: AT THE NULL -- the melody's agreement with its own chords is statistically indistinguishable from pitches drawn at random from the same mode (clearing p95: 0/3; mean percentile 0.661; per cue {'PASS2_FLORES_CACI': 0.9325, 'PASS2_FLORES_REST': 0.5, 'PASS2_FLORES_WATERFALL': 0.55})
round3: AT THE NULL -- the melody's agreement with its own chords is statistically indistinguishable from pitches drawn at random from the same mode (clearing p95: 0/3; mean percentile 0.228; per cue {'PASS5_FLORES_FALLS': 0.0425, 'PASS5_FLORES_ROAD': 0.425, 'PASS5_FLORES_ROUNDS': 0.215})
round4: 1 of 3 cues clear their own null p95 (clearing p95: 1/3; mean percentile 0.781; per cue {'PASS7_FLORES_FALLS': 0.9875, 'PASS7_FLORES_ROAD': 0.4725, 'PASS7_FLORES_ROUNDS': 0.8825})

round	cues	chord_tone_share (preferred src)	same, bar grid	null mean	null percentile	cues above null p95
round2	3	0.495	0.495	0.465	0.661	0/3
round3	3	0.428	0.428	0.470	0.228	0/3
round4	3	0.546	0.506	0.489	0.781	1/3

A percentile near 0.5 means the melody's agreement with its own chords is indistinguishable from pitches drawn at random from the same mode. A percentile of 1.00 with above null p95 means the agreement could not have happened by chance.

The eight axes, per cue

round	cue	src	1 chord_tone_share (prim / all)	2 dissonance_classifiability	3 tension mean / p90 / high-share	4 chords/min, distinct, entropy	5 non-functional	6 supported	7 par 3rds+6ths	8 undecl. parallels
round2	PASS2_FLORES_CACI	bar_grid	0.548 / 0.272	0.390 (133/218 unclassifiable)	1.43 / 4.00 / 0.611	36.38, 7, 3.78 b	0.000	0.922	0.102	113
round2	PASS2_FLORES_REST	bar_grid	0.472 / 0.447	0.544 (31/68 unclassifiable)	1.59 / 4.00 / 0.663	22.50, 7, 3.61 b	0.000	0.920	0.020	80
round2	PASS2_FLORES_WATERFALL	bar_grid	0.467 / 0.467	0.438 (9/16 unclassifiable)	1.45 / 4.00 / 0.275	15.26, 6, 2.95 b	0.000	0.833	0.000	0
round3	PASS5_FLORES_FALLS	bar_grid	0.379 / 0.384	0.395 (66/109 unclassifiable)	1.60 / 4.00 / 0.598	17.33, 6, 3.30 b	0.000	0.975	0.000	73
round3	PASS5_FLORES_ROAD	bar_grid	0.447 / 0.467	0.379 (118/190 unclassifiable)	1.63 / 5.00 / 0.578	28.46, 7, 3.73 b	0.000	0.926	0.000	125
round3	PASS5_FLORES_ROUNDS	bar_grid	0.459 / 0.440	0.358 (147/229 unclassifiable)	1.86 / 4.00 / 0.672	36.40, 7, 3.84 b	0.000	0.905	0.000	51
round4	PASS7_FLORES_FALLS	harmonic_plan	0.580 / 0.551	0.378 (130/209 unclassifiable)	1.59 / 4.00 / 0.627	17.33, 19, 4.83 b	0.346	0.964	0.012	66
round4	PASS7_FLORES_ROAD	harmonic_plan	0.495 / 0.435	0.440 (218/389 unclassifiable)	1.64 / 5.00 / 0.584	13.96, 19, 5.12 b	0.365	0.924	0.007	121
round4	PASS7_FLORES_ROUNDS	harmonic_plan	0.563 / 0.543	0.383 (289/468 unclassifiable)	1.85 / 5.00 / 0.713	17.31, 13, 4.02 b	0.155	0.922	0.001	46

The like-for-like column: every round read off the BAR GRID

Round 4 carries a real harmonic_plan, whose chords include a third whether or not any part voices it; rounds 2 and 3 have only the written voicing. Comparing the preferred sources across rounds would compare two different objects, so this table reads every round off the bar grid, which all three have.

round	cue	chord_tone_share prim (bar grid)	structural tones	null mean	percentile	z
round2	PASS2_FLORES_CACI	0.548	84	0.464	0.932	1.50
round2	PASS2_FLORES_REST	0.472	72	0.479	0.500	-0.13
round2	PASS2_FLORES_WATERFALL	0.467	30	0.451	0.550	0.18
round3	PASS5_FLORES_FALLS	0.379	103	0.467	0.043	-1.86
round3	PASS5_FLORES_ROAD	0.447	179	0.453	0.425	-0.17
round3	PASS5_FLORES_ROUNDS	0.459	146	0.490	0.215	-0.77
round4	PASS7_FLORES_FALLS	0.543	81	0.476	0.988	2.14
round4	PASS7_FLORES_ROAD	0.521	165	0.499	0.472	-0.10
round4	PASS7_FLORES_ROUNDS	0.454	141	0.490	0.882	1.19

The random-melody null, per cue

round	cue	real	null mean	null sd	null p05 / p50 / p95	percentile	z	above p95
round2	PASS2_FLORES_CACI	0.548	0.464	0.056	0.381 / 0.464 / 0.559	0.932	1.50	no
round2	PASS2_FLORES_REST	0.472	0.479	0.054	0.403 / 0.472 / 0.569	0.500	-0.13	no
round2	PASS2_FLORES_WATERFALL	0.467	0.451	0.086	0.300 / 0.467 / 0.600	0.550	0.18	no
round3	PASS5_FLORES_FALLS	0.379	0.467	0.048	0.379 / 0.476 / 0.534	0.043	-1.86	no
round3	PASS5_FLORES_ROAD	0.447	0.453	0.038	0.391 / 0.453 / 0.514	0.425	-0.17	no
round3	PASS5_FLORES_ROUNDS	0.459	0.490	0.041	0.418 / 0.493 / 0.555	0.215	-0.77	no
round4	PASS7_FLORES_FALLS	0.580	0.476	0.048	0.395 / 0.469 / 0.556	0.988	2.14	yes
round4	PASS7_FLORES_ROAD	0.495	0.499	0.047	0.421 / 0.495 / 0.579	0.472	-0.10	no
round4	PASS7_FLORES_ROUNDS	0.563	0.490	0.062	0.394 / 0.493 / 0.592	0.882	1.19	no

The accompaniment's own chord fidelity, split by pattern kind

supported_share asks whether the tune had something under it. This asks whether what was under it is playing the harmony it was handed. Only pad and arpeggio ever read the chord grid (pass2_realise.py:262-288); every other kind is degree-space figuration that was never chord-aware, so near-chance fidelity is expected there. Near-chance fidelity on the two kinds that DO read the grid is the defect named at the top of the module: a voicing written in absolute MIDI is routed through degree_to_midi (pass2_realise.py:425-431) and comes out as a different pitch class.

round	cue	all accompaniment	kinds that READ the grid	kinds that do not
round2	PASS2_FLORES_CACI	0.455	arpeggio 0.555, pad 0.504	kotekan 0.433, ostinato 0.408, pokok 0.415
round2	PASS2_FLORES_REST	0.527	arpeggio 0.579, pad 0.515	drone 1.000, kotekan 0.456, ostinato 0.472, pokok 0.481
round2	PASS2_FLORES_WATERFALL	0.698	arpeggio 0.677, pad 0.682	drone 1.000, pokok 0.531
round3	PASS5_FLORES_FALLS	0.515	pad 0.619	drone 1.000, score 0.427
round3	PASS5_FLORES_ROAD	0.529	arpeggio 0.562, pad 0.523	drone 1.000, ostinato 0.488, pokok 0.487, score 0.480
round3	PASS5_FLORES_ROUNDS	0.461	pad 0.512	counterline 0.000, drone 1.000, ostinato 0.364, pokok 0.467, score 0.418
round4	PASS7_FLORES_FALLS	0.529	none present	drone 0.571, score 0.525
round4	PASS7_FLORES_ROAD	0.565	none present	drone 0.611, ostinato 0.463, pokok 0.457, score 0.573
round4	PASS7_FLORES_ROUNDS	0.559	none present	drone 0.750, ostinato 0.468, pokok 0.497, score 0.556

Round 4 only: does the bar grid agree with the harmonic plan?

cue	bars	grid is a subset of the plan's chord	identical
PASS7_FLORES_FALLS	54	0.537	0.463
PASS7_FLORES_ROAD	108	0.778	0.667
PASS7_FLORES_ROUNDS	124	0.581	0.581

the bar grid is what pad/arpeggio play; the harmonic plan is what the engine declared. A low subset share means the two disagree and the preferred-source number is the more generous of two readings.

Per-axis verdicts

round2 -- PASS2_FLORES_CACI

[FAIL] chord_tone_share_primary = 0.5476 -- target: above the cue's own random-melody null p95 (DERIVED (random-melody null); chord_tone_share is gated against its own random-melody null and never against a fixed consonance quota: docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:63-64,71-73 put the strict ratio of consonance types on the does-not-survive list)
[FAIL] dissonance_classifiability = 0.3899 -- target: 1.0 minus declared exceptions (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:400-403)
[FAIL] undeclared_parallel_perfects = 113 -- target: 0 (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:126-134 (VL-1), counted by harmony.check H01)
[PASS] non_functional_ratio = 0.0 -- target: arrival <= 0.15 (CITED docs/proposals/music/school_grammars/SNES_JRPG_SCHOOL.md:66 (HARM-04))
[report] tension_mean = 1.4348 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] tension_p90 = 4.0 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] high_tension_time_share = 0.6107 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] harmonic_rhythm_per_min = 36.3833 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] distinct_chords = 7 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] chord_change_entropy = 3.7792 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] supported_share = 0.9219 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] parallel_thirds_sixths_share = 0.1021 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)

tier=score; harmony source=bar_grid; sources available=['bar_grid']
15 melodic parts (9 primary) against 15 pitched accompaniment parts
random-melody null: real 0.5476 vs null mean 0.4637 (sd 0.0558), percentile 0.9325, z 1.5033 over 200 seeded draws
NO CORPUS CALIBRATION: the corpus is audio and there is no score for it, so this instrument is a score-side DEFECT DETECTOR and not a taste oracle.

round2 -- PASS2_FLORES_REST

[FAIL] chord_tone_share_primary = 0.4722 -- target: above the cue's own random-melody null p95 (DERIVED (random-melody null); chord_tone_share is gated against its own random-melody null and never against a fixed consonance quota: docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:63-64,71-73 put the strict ratio of consonance types on the does-not-survive list)
[FAIL] dissonance_classifiability = 0.5441 -- target: 1.0 minus declared exceptions (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:400-403)
[FAIL] undeclared_parallel_perfects = 80 -- target: 0 (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:126-134 (VL-1), counted by harmony.check H01)
[FAIL] non_functional_ratio = 0.0 -- target: wonder >= 0.30 (CITED docs/proposals/music/school_grammars/SNES_JRPG_SCHOOL.md:66 (HARM-04))
[report] tension_mean = 1.593 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] tension_p90 = 4.0 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] high_tension_time_share = 0.6635 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] harmonic_rhythm_per_min = 22.5 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] distinct_chords = 7 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] chord_change_entropy = 3.6122 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] supported_share = 0.9196 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] parallel_thirds_sixths_share = 0.0196 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)

tier=score; harmony source=bar_grid; sources available=['bar_grid']
10 melodic parts (8 primary) against 15 pitched accompaniment parts
random-melody null: real 0.4722 vs null mean 0.4792 (sd 0.0541), percentile 0.5, z -0.1284 over 200 seeded draws
NO CORPUS CALIBRATION: the corpus is audio and there is no score for it, so this instrument is a score-side DEFECT DETECTOR and not a taste oracle.

round2 -- PASS2_FLORES_WATERFALL

[FAIL] chord_tone_share_primary = 0.4667 -- target: above the cue's own random-melody null p95 (DERIVED (random-melody null); chord_tone_share is gated against its own random-melody null and never against a fixed consonance quota: docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:63-64,71-73 put the strict ratio of consonance types on the does-not-survive list)
[FAIL] dissonance_classifiability = 0.4375 -- target: 1.0 minus declared exceptions (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:400-403)
[PASS] undeclared_parallel_perfects = 0 -- target: 0 (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:126-134 (VL-1), counted by harmony.check H01)
[FAIL] non_functional_ratio = 0.0 -- target: wonder >= 0.30 (CITED docs/proposals/music/school_grammars/SNES_JRPG_SCHOOL.md:66 (HARM-04))
[report] tension_mean = 1.4513 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] tension_p90 = 4.0 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] high_tension_time_share = 0.275 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] harmonic_rhythm_per_min = 15.2625 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] distinct_chords = 6 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] chord_change_entropy = 2.9514 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] supported_share = 0.8333 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] parallel_thirds_sixths_share = 0.0 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)

tier=score; harmony source=bar_grid; sources available=['bar_grid']
4 melodic parts (4 primary) against 5 pitched accompaniment parts
random-melody null: real 0.4667 vs null mean 0.451 (sd 0.0858), percentile 0.55, z 0.1826 over 200 seeded draws
NO CORPUS CALIBRATION: the corpus is audio and there is no score for it, so this instrument is a score-side DEFECT DETECTOR and not a taste oracle.

round3 -- PASS5_FLORES_FALLS

[FAIL] chord_tone_share_primary = 0.3786 -- target: above the cue's own random-melody null p95 (DERIVED (random-melody null); chord_tone_share is gated against its own random-melody null and never against a fixed consonance quota: docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:63-64,71-73 put the strict ratio of consonance types on the does-not-survive list)
[FAIL] dissonance_classifiability = 0.3945 -- target: 1.0 minus declared exceptions (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:400-403)
[FAIL] undeclared_parallel_perfects = 73 -- target: 0 (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:126-134 (VL-1), counted by harmony.check H01)
[FAIL] non_functional_ratio = 0.0 -- target: wonder >= 0.30 (CITED docs/proposals/music/school_grammars/SNES_JRPG_SCHOOL.md:66 (HARM-04))
[report] tension_mean = 1.5954 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] tension_p90 = 4.0 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] high_tension_time_share = 0.5977 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] harmonic_rhythm_per_min = 17.3333 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] distinct_chords = 6 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] chord_change_entropy = 3.2975 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] supported_share = 0.9751 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] parallel_thirds_sixths_share = 0.0 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)

tier=score; harmony source=bar_grid; sources available=['bar_grid']
6 melodic parts (4 primary) against 8 pitched accompaniment parts
random-melody null: real 0.3786 vs null mean 0.4671 (sd 0.0475), percentile 0.0425, z -1.8624 over 200 seeded draws
NO CORPUS CALIBRATION: the corpus is audio and there is no score for it, so this instrument is a score-side DEFECT DETECTOR and not a taste oracle.

round3 -- PASS5_FLORES_ROAD

[FAIL] chord_tone_share_primary = 0.4469 -- target: above the cue's own random-melody null p95 (DERIVED (random-melody null); chord_tone_share is gated against its own random-melody null and never against a fixed consonance quota: docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:63-64,71-73 put the strict ratio of consonance types on the does-not-survive list)
[FAIL] dissonance_classifiability = 0.3789 -- target: 1.0 minus declared exceptions (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:400-403)
[FAIL] undeclared_parallel_perfects = 125 -- target: 0 (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:126-134 (VL-1), counted by harmony.check H01)
[FAIL] non_functional_ratio = 0.0 -- target: wonder >= 0.30 (CITED docs/proposals/music/school_grammars/SNES_JRPG_SCHOOL.md:66 (HARM-04))
[report] tension_mean = 1.6327 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] tension_p90 = 5.0 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] high_tension_time_share = 0.578 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] harmonic_rhythm_per_min = 28.463 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] distinct_chords = 7 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] chord_change_entropy = 3.7252 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] supported_share = 0.9259 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] parallel_thirds_sixths_share = 0.0 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)

tier=score; harmony source=bar_grid; sources available=['bar_grid']
9 melodic parts (7 primary) against 15 pitched accompaniment parts
random-melody null: real 0.4469 vs null mean 0.4532 (sd 0.0378), percentile 0.425, z -0.1654 over 200 seeded draws
NO CORPUS CALIBRATION: the corpus is audio and there is no score for it, so this instrument is a score-side DEFECT DETECTOR and not a taste oracle.

round3 -- PASS5_FLORES_ROUNDS

[FAIL] chord_tone_share_primary = 0.4589 -- target: above the cue's own random-melody null p95 (DERIVED (random-melody null); chord_tone_share is gated against its own random-melody null and never against a fixed consonance quota: docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:63-64,71-73 put the strict ratio of consonance types on the does-not-survive list)
[FAIL] dissonance_classifiability = 0.3581 -- target: 1.0 minus declared exceptions (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:400-403)
[FAIL] undeclared_parallel_perfects = 51 -- target: 0 (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:126-134 (VL-1), counted by harmony.check H01)
[PASS] non_functional_ratio = 0.0 -- target: arrival <= 0.15 (CITED docs/proposals/music/school_grammars/SNES_JRPG_SCHOOL.md:66 (HARM-04))
[report] tension_mean = 1.861 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] tension_p90 = 4.0 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] high_tension_time_share = 0.6719 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] harmonic_rhythm_per_min = 36.4032 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] distinct_chords = 7 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] chord_change_entropy = 3.8388 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] supported_share = 0.9049 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] parallel_thirds_sixths_share = 0.0 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)

tier=score; harmony source=bar_grid; sources available=['bar_grid']
9 melodic parts (7 primary) against 16 pitched accompaniment parts
random-melody null: real 0.4589 vs null mean 0.4903 (sd 0.0408), percentile 0.215, z -0.7701 over 200 seeded draws
NO CORPUS CALIBRATION: the corpus is audio and there is no score for it, so this instrument is a score-side DEFECT DETECTOR and not a taste oracle.

round4 -- PASS7_FLORES_FALLS

[PASS] chord_tone_share_primary = 0.5802 -- target: above the cue's own random-melody null p95 (DERIVED (random-melody null); chord_tone_share is gated against its own random-melody null and never against a fixed consonance quota: docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:63-64,71-73 put the strict ratio of consonance types on the does-not-survive list)
[FAIL] dissonance_classifiability = 0.378 -- target: 1.0 minus declared exceptions (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:400-403)
[FAIL] undeclared_parallel_perfects = 66 -- target: 0 (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:126-134 (VL-1), counted by harmony.check H01)
[PASS] non_functional_ratio = 0.3462 -- target: wonder >= 0.30 (CITED docs/proposals/music/school_grammars/SNES_JRPG_SCHOOL.md:66 (HARM-04))
[report] tension_mean = 1.5862 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] tension_p90 = 4.0 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] high_tension_time_share = 0.627 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] harmonic_rhythm_per_min = 17.3333 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] distinct_chords = 19 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] chord_change_entropy = 4.8296 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] supported_share = 0.9635 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] parallel_thirds_sixths_share = 0.012 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)

tier=score; harmony source=harmonic_plan; sources available=['bar_grid', 'harmonic_plan']
6 melodic parts (4 primary) against 8 pitched accompaniment parts
random-melody null: real 0.5802 vs null mean 0.4765 (sd 0.0484), percentile 0.9875, z 2.1419 over 200 seeded draws
NO CORPUS CALIBRATION: the corpus is audio and there is no score for it, so this instrument is a score-side DEFECT DETECTOR and not a taste oracle.

round4 -- PASS7_FLORES_ROAD

[FAIL] chord_tone_share_primary = 0.4947 -- target: above the cue's own random-melody null p95 (DERIVED (random-melody null); chord_tone_share is gated against its own random-melody null and never against a fixed consonance quota: docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:63-64,71-73 put the strict ratio of consonance types on the does-not-survive list)
[FAIL] dissonance_classifiability = 0.4396 -- target: 1.0 minus declared exceptions (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:400-403)
[FAIL] undeclared_parallel_perfects = 121 -- target: 0 (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:126-134 (VL-1), counted by harmony.check H01)
[PASS] non_functional_ratio = 0.3654 -- target: wonder >= 0.30 (CITED docs/proposals/music/school_grammars/SNES_JRPG_SCHOOL.md:66 (HARM-04))
[report] tension_mean = 1.6369 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] tension_p90 = 5.0 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] high_tension_time_share = 0.5844 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] harmonic_rhythm_per_min = 13.963 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] distinct_chords = 19 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] chord_change_entropy = 5.1184 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] supported_share = 0.9245 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] parallel_thirds_sixths_share = 0.0067 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)

tier=score; harmony source=harmonic_plan; sources available=['bar_grid', 'harmonic_plan']
10 melodic parts (7 primary) against 15 pitched accompaniment parts
random-melody null: real 0.4947 vs null mean 0.4995 (sd 0.047), percentile 0.4725, z -0.102 over 200 seeded draws
NO CORPUS CALIBRATION: the corpus is audio and there is no score for it, so this instrument is a score-side DEFECT DETECTOR and not a taste oracle.

round4 -- PASS7_FLORES_ROUNDS

[FAIL] chord_tone_share_primary = 0.5634 -- target: above the cue's own random-melody null p95 (DERIVED (random-melody null); chord_tone_share is gated against its own random-melody null and never against a fixed consonance quota: docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:63-64,71-73 put the strict ratio of consonance types on the does-not-survive list)
[FAIL] dissonance_classifiability = 0.3825 -- target: 1.0 minus declared exceptions (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:400-403)
[FAIL] undeclared_parallel_perfects = 46 -- target: 0 (CITED docs/proposals/music/craft_research/COUNTERPOINT_AND_VOICE_LEADING.md:126-134 (VL-1), counted by harmony.check H01)
[FAIL] non_functional_ratio = 0.1552 -- target: arrival <= 0.15 (CITED docs/proposals/music/school_grammars/SNES_JRPG_SCHOOL.md:66 (HARM-04))
[report] tension_mean = 1.8465 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] tension_p90 = 5.0 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] high_tension_time_share = 0.7129 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] harmonic_rhythm_per_min = 17.3065 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] distinct_chords = 13 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] chord_change_entropy = 4.0243 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] supported_share = 0.9223 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)
[report] parallel_thirds_sixths_share = 0.0007 -- target: REPORT ONLY -- no corpus calibration exists for a score axis (DERIVED)

tier=score; harmony source=harmonic_plan; sources available=['bar_grid', 'harmonic_plan']
10 melodic parts (7 primary) against 16 pitched accompaniment parts
random-melody null: real 0.5634 vs null mean 0.4898 (sd 0.062), percentile 0.8825, z 1.1875 over 200 seeded draws
NO CORPUS CALIBRATION: the corpus is audio and there is no score for it, so this instrument is a score-side DEFECT DETECTOR and not a taste oracle.

Generated by harness/site/structure_site.py — the URL path is the repo path. review root