FDT Bell v20: the projection settlement
=======================================
Chart and status (core v6.7): the helix chart is the doubled-angle circle of the v19 cylinder readout;
the records A = sign cos(phi + 2 beta_A - a), B = sign cos(phi + pi + 2 beta_B - b) are unchanged.
The shared phase carries the projection measure of the host clamp, rho(phi | a) = |cos(phi - a)|/4:
the clamp reads the helix in projection (I-4) and the pair's record is one system-wide settlement
(0.1a) whose mode geometry is set by the clamps (P-2).  The measure is set by a clamp; a
clamp-independent measure keeps |S| <= 2 as arithmetic and runs beside every settlement number.
No number of this report's settlement columns is read from the Born layer; the Born-layer columns
are the v19 values, recomputed in this run, that the settlement is compared against.
field seed=314159; settlement seed=271828; histories=1536; step=0.002 m; records per setting: ideal=1000000, scenario=250000, detector=20000, sphere=2000000

[1] The joint impact distribution, exact arc integration (D-B1)
  p(phi_A, phi_B | a, b) = delta(phi_B - phi_A - pi) rho(phi_A | host clamp); sign readout at each wing
  scan: Delta = a - b - 2(beta_A - beta_B) on 1441 points of [-2 pi, 2 pi]
  measure             host     max|E + cos Delta|  max|P(A=+) - 1/2|  max|P(B=+) - 1/2|  max|sum - 1|
  projection |cos|/4  A        4.441e-16           1.110e-16          1.110e-16          2.220e-16
  projection |cos|/4  B        5.551e-16           1.110e-16          1.110e-16          2.220e-16
  projection |cos|/4  either   3.331e-16           1.110e-16          1.110e-16          2.220e-16
  host invariance: max cell difference between host A and a host mixture (weights 0, 1/4, 1/2, 3/4, 1) = 1.665e-16
  uniform 1/(2 pi), clamp-independent: max|E - (-1 + 2|Delta|/pi)| = 3.331e-16; max|E + (2/pi) arcsin cos Delta| = 1.887e-15
  CHSH at the v19 clamps a = (0, pi/2), b = (-3pi/4, 3pi/4): projection S = 2.828427125 (2 sqrt 2 = 2.828427125); uniform S = 2.000000000; ratio = 1.414213562 = K_G(2)
  ceiling over 331776 clamp quadruples (15 degree grid) and four facets: projection max|S| = 2.828427125; uniform max|S| = 2.000000000

[2] Every record kept: event-by-event CHSH at the v19 clamps (1000000 records per setting)
  measure, host         E00       E01       E10       E11       S +/- SE              S exact      kept/declared      max cell z
  projection, host A   +0.70727  +0.70711  +0.70748  -0.70688  2.828730 +/- 0.001414  2.828427     4000000/4000000  1.26
  projection, host B   +0.70814  +0.70729  +0.70718  -0.70687  2.829470 +/- 0.001414  2.828427     4000000/4000000  1.84
  projection, either   +0.70589  +0.70676  +0.70789  -0.70709  2.827638 +/- 0.001414  2.828427     4000000/4000000  2.08
  uniform (control)    +0.49841  +0.50082  +0.50001  -0.50095  2.000188 +/- 0.001732  2.000000     4000000/4000000  2.30
  single-wing records: max |P(+) - 1/2|, Alice's change under Bob's clamp, Bob's under Alice's (sampled)
    projection, host A   0.00043   0.00039   0.00047
    projection, host B   0.00120   0.00184   0.00111
    projection, either   0.00074   0.00033   0.00096
    uniform (control)    0.00092   0.00134   0.00040
  counterfactual CHSH on one phase (all four clamp positions read on the same phi): support projection, host A: [2]; projection, host B: [2]; projection, either: [2]; uniform (control): [2]
  The arithmetic bound holds phase by phase in every row.  The settlement's four correlators are
  averages over four measures, one per host setting; the uniform row averages over one.

[3] The clamp route priced (D-B2)
  Hall's measure M = sup int |rho_xy - rho_x'y'| dphi on the v19 clamps; M_1 varies A's clamp, M_2 varies B's
    host A               M = 0.828427125   M_1 = 0.828427125   M_2 = 0.000000000   F = 1 - M/2 = 0.585786
    host B               M = 0.828427125   M_1 = 0.000000000   M_2 = 0.828427125   F = 1 - M/2 = 0.585786
    either, weight 1/2   M = 0.613125930   M_1 = 0.414213562   M_2 = 0.414213562   F = 1 - M/2 = 0.693437
  one host, all clamp pairs: sup M_1 = 0.828427125 at clamps 90.0 degrees apart; closed form 2 sqrt 2 - 2 = 0.828427125
  one-sided relaxed bound S <= 2 + M_1: S = 2.828427125; 2 + M_1 = 2.828427125; slack = 1.239e-12; int (rho_0 - rho_1) B_0 B_1 dphi = 0.828427125 = M_1
  mutual information between the phase and the host clamp: 0.201752 bits (two equiprobable clamps); 0.208801 bits (clamp uniform on the circle; closed form 0.208801)
  reference (Hall 2010): a measure free to depend on both clamps obeys S <= 2 + 3M, so 2 sqrt 2 needs M >= 0.276142 and S = 4 needs M = 2/3.
  ceiling: sup M_1 over all clamp pairs is 0.828427125, so S <= 2 + M_1 <= 2 sqrt 2 on every quartet;
  the Tsirelson value is the projection measure's own ceiling, met at clamps a quarter turn apart (2.828427125 on the grid of [1]).
  A measure free to depend on both clamps reaches 4 at M = 2/3; the projection measure stops at 2 sqrt 2.
  prediction: an all-record CHSH above 2 sqrt 2, or a unit-visibility kernel other than -cos Delta,
  falsifies the projection measure.

[4] The same weight by selection (1000000 records per setting)
  uniform phase; clamp A accepts with probability |cos(phi - a)| (reads its own clamp only)
    accepted-pair S = 2.828682; wing efficiency = 0.636600 (2/pi = 0.636620); S with nulls as 0 = 1.800740; S with nulls as -1 = 1.801084
    settlement: S = 2.827638 with 4000000/4000000 records kept.
  Selection reaches the measure |cos|/4 by dropping records and stays under 2 on all records;
  the settlement carries the same measure on every record.

[5] The v19 chain with the settlement in place of the Born layer (1536 histories; 250000 records per setting)
  scenario                  S Born layer  S settlement  max cell diff  host gap   S sampled +/- SE        z      S uniform
  uniform_fluctuations      2.828427125   2.828427125   7.161e-15      2.776e-16  2.825576 +/- 0.002828  -1.01  2.000000
  matched                   2.828427125   2.828427125   7.216e-15      2.776e-16  2.828008 +/- 0.002828  -0.15  2.000000
  independent               1.750213217   1.750213217   6.661e-16      5.551e-17  1.753728 +/- 0.003596  +0.98  1.364052
  partial_shared            1.882046657   1.882046657   2.220e-16      1.110e-16  1.886984 +/- 0.003530  +1.40  1.447986
  shared_unequal_baselines  1.450123196   1.450123196   2.776e-16      5.551e-17  1.449224 +/- 0.003727  -0.24  1.127653
  baseline_low              2.563461298   2.563461298   3.331e-16      1.110e-16  2.562912 +/- 0.003070  -0.18  1.887369
  baseline_high             0.651282774   0.651282774   3.053e-16      5.551e-17  0.655144 +/- 0.003947  +0.98  0.529043
  ordered_regions           2.828427125   2.828427125   7.161e-15      2.776e-16  2.823728 +/- 0.002828  -1.66  2.000000
  reordered_regions         2.759170303   2.759170303   5.551e-16      1.110e-16  2.756456 +/- 0.002893  -0.94  1.961892
  S settlement: exact arc integration on each history's clamps, averaged over the histories.
  max cell diff: the largest difference over the sixteen (setting, record) cells against the Born table.

[6] Detector experiment, event by event (independent scenario; 85% pair source; no_click -> -1; 20000 trials per setting)
  all-trial S (the Bell ledger)      Born layer   settlement   sampled +/- SE
    ideal                            1.787681     1.787681     1.787400 +/- 0.012535
    fixed_loss_and_readout           1.482804     1.482804     1.506600 +/- 0.013008
    history_correlated_collection    1.223112     1.223112     1.228200 +/- 0.012622
  coincidence-conditioned S          Born layer   settlement   sampled +/- SE
    ideal                            1.750213     1.750213     1.756445 +/- 0.013792
    fixed_loss_and_readout           1.660082     1.660082     1.662293 +/- 0.015782
    history_correlated_collection    2.316859     2.316859     2.323252 +/- 0.026982
  nine-cell tables                   max|settlement - Born|  |sum - 1|   A marginal change  B marginal change  chi2/dof of counts
    ideal                            4.441e-16               1.332e-15   7.772e-16          7.772e-16          7.28/16
    fixed_loss_and_readout           1.221e-15               8.882e-16   1.110e-15          1.332e-15          33.66/32
    history_correlated_collection    9.437e-16               6.661e-16   6.661e-16          7.216e-16          20.58/32
  marginal change: a wing's complete three-outcome marginal under a change of the remote clamp.
  observed coincidences (00, 01, 10, 11): ideal: 17048, 17003, 16990, 16939; fixed_loss_and_readout: 13303, 13261, 13270, 13319; history_correlated_collection: 3626, 3641, 3632, 3703
  All nine count cells and both probability tables per setting: settlement_ledger_v20.json

[7] The Hensen (2015) spin axes, without data
  uniform half-cylinder S = 2.000000; projection settlement S = 2.818997; -cos kernel at the same axes = 2.818997

[8] The sphere of the helix chart: arbitrary analyzer axes on the disc measure (2000000 records per pair)
  rho(lambda | s) = |s . lambda|/(2 pi): uniform on the disc normal to the settled axis, lifted to the sphere
  a.b         E sampled +/- SE        -a.b        z      P(A=+)    P(B=+)
  +0.000000   -0.000128 +/- 0.000707   +0.000000   -0.18   0.50007   0.50001
  +0.511959   -0.512322 +/- 0.000607   -0.511959   -0.60   0.49932   0.50026
  +0.192450   -0.192400 +/- 0.000694   -0.192450   +0.07   0.49965   0.50031
  +0.994907   -0.994952 +/- 0.000071   -0.994907   -0.63   0.50015   0.49986
  great-circle section of the disc measure against |cos|/4: Kolmogorov distance = 0.000730 (1% critical value 0.001153)
  Horodecki-optimal axes, independent scenario, through the per-history transport: S sampled = 2.353295 +/- 0.001082; 2 sqrt(1 + |gamma|^2) = 2.352637; z = +0.61

Reproducibility
  field draws use seed 314159 and the v19 named streams; every settlement block draws from its own
  substream of seed 271828; the report and the ledger are byte-reproducible under the same
  seeds and runtime.

Decision checks
PASS: the sign readout on the projection measure returns E = -cos Delta to 1e-12 on the whole scan, for every host rule.
PASS: both single-wing marginals are 1/2 to 1e-12 at every scan point and in every scenario and setting.
PASS: the table is the same whichever clamp hosts the settlement (1e-12), on the scan and in every scenario.
PASS: the clamp-independent uniform measure returns the triangle, S = 2 at the v19 clamps, |S| <= 2 on the quadruple grid, and the counterfactual CHSH support lies in {-2, +2} at every sampled phase.
PASS: the projection measure gives S = 2 sqrt 2 at the v19 clamps and max|S| = 2 sqrt 2 on the quadruple grid.
PASS: M_1 = 2 sqrt 2 - 2 and S = 2 + M_1 to 1e-9 (midpoint quadrature); the other host's dependence is zero.
PASS: in all nine v19 scenarios the settlement table equals the Born-layer table cell by cell to 1e-12.
PASS: every nine-cell settlement table is normalized, no-signalling and equal to the Born-layer table to 1e-12.
PASS: every declared trial is recorded in every sampled block.
PASS: at the Hensen axes the settlement returns the -cos kernel value and the uniform half-cylinder returns 2.
PASS: the circle clamps of the settlement are the v19 Stokes analyzer axes (sin a, 0, cos a).

Sampling agreement at the declared sample sizes (reported, not gated)
  largest |z| of a sampled CHSH against its exact settlement value: scenarios 1.66; detector all-trial 1.83; detector coincidence 0.45; sphere pairs 0.63

INTERNAL IMPLEMENTATION AUDIT: PASS
