Stonehenge | S8 | v1.39 | 1
Supplementary Material S8
Stellar fields without constellation boundaries
Lucas Giovani Ribeiro and Marcos André Simonssini
Companion to Stonehenge revision v1.39 | 29 October 2026
The complete nominal Hipparcos J2000 census ranks Cygnus first among 19,939 fields under the fixed 
angular target (7°,18°,4°), without constellation boundaries or a four-arm entry filter. The matching free-scale 
control gives rank 68 with a nearby Cygnus epoch. Section 9 reports that extension. Sections 1–3 preserve 
the earlier cap-centred four-arm ratio panels and their targeted kinematic variants, whose robustness does 
not automatically extend to the new complete fixed-scale ranking. Sections 4–7 retain invariance and 
geometric-ratio diagnostics, including unsuccessful controls. All analyses are retrospective.
S8.1 Field enumeration and role selection
Use every entry in the archived V≤6 catalogue as the centre of a spherical cap. Increase its radius through 
every stellar entry event and retain each distinct set of the five brightest enclosed entries while the centre 
itself remains among those five. Apparent-magnitude ties are broken by catalogue identifier. Once the centre 
is displaced from the five brightest, enlarging the cap cannot restore it, so enumeration for that centre can 
stop. No cap radius is optimized against a clock residual.
For the original ratio panels in sections 2–3, retain a quintet only if its unique angular-distance minimum 
spanning tree consists of four edges incident on the cap centre h. For every pair of leaves i and j, the strict 
test is d(i,j)>max[d(h,i),d(h,j)], with implementation tolerance 10 ¹  degree. Kruskal construction and the strict
⁻ ⁰
 
cycle-margin test agree. The five-brightest membership at each retained interval midpoint is checked against 
the entire catalogue. This is a four-arm topological condition, not a right-angled visual cross. Section 9 
retains every generated quintet before this filter.
The first endpoint is the brightest non-hub star. The second endpoint is the remaining star most opposite in 
tangent bearing at the hub. Their two adjacent intervals define channels 1 and 2; the residual unordered pair 
defines channel 3. This is the same geometric role rule as S6 and the separate stellar manuscript [main 
reference 47, section 2.9], here restricted to star-shaped trees. Constellation labels, archaeological 
chronology, the horizon, the spatial tuple and 7:18:4 do not enter candidate generation. The group called 
Cygnus is identified by its five HIP identifiers only when results are reported.
This family was selected after the Cygnus geometry and previous timing results were known. It is not a 
preregistered discovery, a multiplicity-calibrated significance test or an exhaustive enumeration of all five-
star subsets. Unlike S7, it does not admit regional quintets centred only by a medoid. Thus its 256 nominal 
candidates are not a simple expansion of the earlier 88-group comparison; both are retained. The complete 
extension in section 9 also admits medoids, while retaining the cap-generation rule.
Cygnus roles are Deneb (HIP 102098), Sadr (100453), Albireo (95947), and the unordered pair Aljanah 
(102488) and Fawaris (97165). Reversing that final pair leaves the unsigned third interval unchanged. The 
centre is one of the five stars, so there are four nonzero hub distances.

Stonehenge | S8 | v1.39 | 2
S8.2 Nominal panels and timing results
Hipparcos supplies 5,041 and HYG 5,070 archived entries with V≤6. Groups are constructed independently 
at J2000 and 3000 BCE, keeping nominal apparent magnitudes fixed. Hipparcos uses normalized angular 
proper motion from J1991.25. HYG uses linear spatial motion from J2000, reconstructed from archived 
positions, distances, proper motions and radial velocities; missing radial velocity is zero. These catalogues 
have shared provenance and are not independent observations.
Table S8.1. Every nominal field panel. R is the horizon-constrained runner-up error divided by Cygnus 
error.
Catalogue / geometry
Quintets 
inspected
Four-arm 
fields
Horizon fields
Rank global / local
R
Hipparcos / J2000
19,939
256
232
1 / 1
1.971
Hipparcos / 3000 BCE
19,904
267
243
1 / 1
3.890
HYG / J2000
19,939
255
231
1 / 1
1.679
HYG / 3000 BCE
19,923
261
237
1 / 1
3.888
Every retained group is searched over astronomical years −8000 to +2000, independently of its group-
construction epoch. BCE=1−astronomical year; positive astronomical years are reported as CE. The score is 
E=sd(d₁/7,d₂/18,d₃/4)/mean(d₁/7,d₂/18,d₃/4), using population standard deviation. Absolute shortest right-
ascension differences are measured relative to the equator of date. The common angular unit cancels. All 
bracketed minima and endpoints on both 0.5-year and 1-year grids are refined. Numerical agreement is not 
an interval certificate of global optimization.
Hipparcos gives Cygnus E=0.006490264894 at 2985.482490 BCE; HYG gives E=0.006490067261 at 
2986.051275 BCE. Both construction epochs recover the same roles. The Hipparcos J2000 cap around 
Sadr retains the same five brightest members for radii approximately 16.238753–20.283777°. At 3000 BCE 
its interval is 16.225103–20.641986°. These intervals demonstrate membership over a range of radii, rather 
than a single radius chosen from temporal fit.
The nominal global runner-up is centred on HIP 109268. In Hipparcos it reaches E=0.006590907555 at 
1582.263739 CE, only 1.5507% above Cygnus. This small global margin must be reported alongside first 
place. The regional name associated with this star is not an input to generation.
The necessary horizon condition is hᵢ=90°−|51.18°−δᵢ|>0 for all five members at the candidate epoch. Every 
admissible segment is found, its boundary refined and E reoptimized within it, retaining boundary infima. It is 
not enough to remove an invisible unrestricted optimum. The condition addresses each star’s ability to rise, 
not simultaneous darkness, terrain, atmosphere or an actual observing occasion.
In every nominal panel, Cygnus remains first at nested ceilings V≤4, 5 and 6, with and without the horizon 
condition. In Hipparcos J2000 these have 22, 66 and 256 candidates, of which 19, 59 and 232 have 
admissible epochs. The lower ceilings are obtained by restricting the faintest member: stars dimmer than 
that member cannot displace a brighter top-five set.

Stonehenge | S8 | v1.39 | 3
S8.3 Targeted kinematic sensitivity and comparison scope
For each of the six archived S6 state variants, replace all five Cygnus states in the HYG catalogue and 
repeat the full J2000 field census. Refit every new field and every field containing any changed star. Reuse 
an earlier score only for unchanged member states and roles. Four baseline fields contain a Cygnus 
member; this procedure therefore does more than revise one score row. Each variant returns 255 fields, with 
231 having an admissible epoch.
Table S8.2. Six complete recensus variants. E is Cygnus relative dispersion; R is the local 
runner-up/Cygnus error ratio.
Variant
Global 
rank
Local rank
E (%)
Epoch (BCE)
R
Archived HYG states
1
1
0.649012
2986.053613
1.679
Hip2 proper motion
1
1
0.653522
2986.973039
1.667
Hip2 radial velocity
1
1
0.653521
2986.973037
1.667
Hip2 with HGCA
2
1
0.661960
2985.499905
1.646
Albireo systemic motion
2
1
0.690059
2984.946520
1.579
Deneb 802 pc sensitivity
2
1
0.690059
2984.946533
1.579
The local first-place result survives all six targeted variants, with the smallest margin 1.5789. Global first 
place does not: Cygnus is second in the final three rows. The range 2984.95–2986.97 BCE measures 
sensitivity to these state substitutions under a fixed objective and background catalogue. It is not a 
confidence interval and does not replace the broader loss sensitivities. S6 provides the state definitions and 
provenance; these are six targeted catalogues, not six independent complete surveys. The archived HYG 
state row differs slightly from the nominal HYG census because its Cygnus states come from the pinned S6 
state file.
Table S8.3. Candidate families remain distinct
Family
Choices retained
Cygnus result
S5 fixed-unit roles
5,280 regional role classes; original 
angular scale
First near 2965 BCE
S5 free-scale roles
Same 5,280 classes; relative-dispersion 
loss
39th near 2985 BCE
S7 geometric roles
88 regional quintets; tree hub or medoid
Second globally; first among 71 horizon-
admissible groups
S8 geometric fields
No regional boundary; star-centred cap 
and four-arm tree
First in four nominal panels; local first in all six 
state variants
The field census eliminates regional boundaries as a prerequisite for this recovery. It retains the count map, 
ordered target, five-member field definition, brightness and role rules, loss, astrometry and search window. 
Its candidate counts and error ratios are descriptive comparisons, not historical odds. Tucana’s earlier 
regional result remains valid under its own rule; the new field family is not used to erase it.

Stonehenge | S8 | v1.39 | 4
S8.4 Coordinate invariance and the relative loss
The preceding invariance audit concerns the S7 assignments and supplies an algebraic interpretation 
applicable to the same intervals in S8. Let uᵢ and u  be unit stellar directions and k the unit physical terrestrial
ⱼ
 
pole at the epoch, all expressed in one common frame. Define the scalar a=uᵢ·u −(uᵢ·k)(u ·k) and b=|
ⱼ
ⱼ
k·(uᵢ×u )|. Then dᵢ=atan2(b,a) is the absolute shortest longitude interval about k, between zero and π. This is
ⱼ
ⱼ
 
the right-ascension interval, not the great-circle separation. If an equatorial projection vanishes, longitude is 
undefined; no tested input encountered this case.
A common orthogonal change of basis of uᵢ, u  and k preserves a and b, including reflections. The formula 
ⱼ
therefore requires neither a right-ascension zero nor chosen equatorial x and y axes. It retains the physical 
pole and its precession. The audited implementation agrees with direct right ascension to approximately 
1.65×10 ¹¹ degree over all 88 assignments and 20,001 epochs. Twenty-four orthogonal basis changes at 
⁻
eleven epochs per group agree to approximately 5.02×10 ¹¹ degree. These checks verify the implementation
⁻
 
of an algebraic invariance already implicit in right-ascension differences.
A fitted common scale
Write qᵢ=dᵢ/wᵢ with w=(7,18,4), and relative residual rᵢ=sqᵢ−1 for a positive common inverse scale s. 
Minimizing the mean squared residual gives s*=Σqᵢ/Σqᵢ². Its minimum is F₂²=1−mean(q)²/mean(q²). Since 
E²=mean(q²)/mean(q)²−1, it follows that F₂=E/√(1+E²). This transformation is strictly increasing, so every 
ranking and epoch minimizer is unchanged. Numerical disagreement between the two forms is at most 
3.33×10 ¹⁶ in the archived test.
⁻
This establishes that the chosen dispersion is equivalent to fitting a common inverse scale with a quadratic 
relative-residual loss. It does not specify a probabilistic measurement-error model. Multiplying all angles by a 
positive constant changes s but leaves E and F₂ unchanged. The fixed target in degrees and the earlier 
target in sidereal minutes instead differ only by unit conversion; freeing scale changes the objective and can 
change rankings, as recorded in S5–S7.
Conditional norm and weight bounds in the regional panel
For each of the 71 horizon-admissible S7 groups, define U  as the infimum over admissible epoch and 
ₚ
positive scale of the p-norm of r. The archived numerical minima give F₂(Cygnus)=0.006490128202 and 
F₂(Norma)=0.028000182275, with Norma the runner-up and R=4.314272600. In dimension three, norm 
equivalence bounds the competitor/Cygnus ratio below by R/√3 for all 1≤p≤∞. This is greater than one, 
preserving first place throughout that family, conditional on the stated numerical minima.
For positive common channel weights λᵢ summing to one, √(3λmin)F₂≤Fλ≤√(3λmax)F₂, after minimizing over 
the same domain. A sufficient condition preserving first place is λmax/λmin<R²=18.612948070. A ratio at 
most 18 therefore suffices. This excludes zero weights, group-specific weights and retrospective choices of 
a different weight vector per competitor. These margins belong to the old regional panel; they are not 
transferred to the new field family.

Stonehenge | S8 | v1.39 | 5
S8.5 Loss controls and continuous spatial reconstruction
Table S8.4. Direct loss reoptimization in the 71-group horizon-admissible regional panel
Relative-residual loss
Cygnus rank
Cygnus optimum (BCE)
Mean absolute residual L1
1
2973.112856
Root mean square residual L2
1
2985.482485
Maximum absolute residual Linf
1
2997.928665
All three sampled-loss optima lie in the cited 3000–2920 BCE initial-enclosure interval. The norm bound 
concerns rank, not constancy of epoch for every p. Numerical minima agree across 0.5-year and 1-year 
grids with local refinement; they are not interval-certified global optima. The epoch span in Table S8.4 is a 
sensitivity result, not an archaeological uncertainty distribution.
Releasing the number-to-interval order
Allow every group all six permutations of 7:18:4, reoptimizing epoch and scale under the same horizon rule. 
There are 528 group-order combinations, 426 admissible. Rank each group by its best ordering. Lyra 
becomes first with 18:4:7, E=0.002880093148 at 1766.714456 CE. Cygnus is second and retains 7:18:4, 
E=0.006490264894 near 2985.482481 BCE. No chronological filter is added to remove Lyra. This exposes 
the role of the ordered map; it does not invalidate the result when that map is retained. The complete 
permutation control belongs to the regional family and has not been repeated as a full field-family 
permutation census.
Removing the ideal triangle and Aubrey calibration
Use the published side means x=33.435 m and y=80.095 m, retain the rectangular hypothesis and calculate 
z=√(x²+y²)=86.793480458 m. The diagonal is derived rather than a third observation. Apply the same 
constructive operations to predict the four interior dimensions without refitting their targets.
Table S8.5. Continuous exterior construction and held-out interior summaries
Interior role
Construction
Output (m)
Published (m)
Trilithon span
(x+y−z)/2
13.368260
13.7
Sarsen outer diameter
x
33.435000
33.0
Z-hole diameter
y/2
40.047500
39.0
Y-hole diameter
z−x
53.358480
54.0
The root-mean-square relative discrepancy is 2.014104407%, close to the original 2.02%. This preserves 
dimensional agreement without imposing an integer triangle or an Aubrey diameter. It uses two exterior 
measures to set shape and scale, whereas the original fixed shape used one estimated scale. The slight 
reduction is not a statistical improvement under equal parameter counts. Rounded inputs, feature 
boundaries and retrospective assignments remain shared.
With r=y/x, dividing every output by x gives [(1+r−√(1+r²))/2, 1, r/2, √(1+r²)−1]. This is a dimensionless 
expression of the same continuous construction. It does not add independent observations or establish 
historical transmission.

Stonehenge | S8 | v1.39 | 6
S8.6 Spherical axis and centre ratios
Use great-circle distance d(u,v)=atan2(‖u×v‖,u·v). Define a=d(Deneb,Sadr), b=d(Sadr,Albireo), 
c=d(Sadr,Fawaris), e=d(Sadr,Aljanah), ℓ=d(Deneb,Albireo) and τ=d(Fawaris,Aljanah). These symbols are 
local to this section and do not denote inventory counts. The archived code uses A,B,C,D,L,T for the same 
six lengths. These distances do not require the terrestrial pole. Common rotations, including a change of 
precession frame, preserve them; their time dependence in the model comes from stellar motion.
Table S8.6. Great-circle geometry at the imported ratio optimum 2985.482485 BCE
Quantity
Value
a, b, c, e (degrees)
6.131807, 16.225142, 8.406342, 8.010386
ℓ, τ (degrees)
22.270452, 16.219591
b/τ
1.000342250
ℓ/τ
1.373058775
Normalized radii a, b, c, e
0.158143542, 0.418457669, 0.216805382, 0.206593407
The Sadr–Albireo arm exceeds the transverse width by 0.034225%. This near equality was noticed while 
inspecting the distances; it was not a prediction made before examining Cygnus. Apply the same |log(b/τ)| 
score to all 88 geometric regional groups at the same imported epoch. Cygnus is first, with runner-up 
Capricornus error 37.684 times larger. It remains first at every half-year sample in 3000–2920 BCE, with 
minimum runner-up ratio 33.191. At J2000 it is second. These ranks condition on the chosen relation and 
common epoch.
Two adverse controls delimit that finding. Allowing each group its closest equality among all 15 pairs of the 
six distances puts Cygnus second at the imported epoch and eighth at J2000. Allowing each group to 
choose an epoch gives exact b/τ=1 crossings in Cancer at 6601.718 BCE, Cygnus at 3526.377 BCE, Cetus 
at 1058.161 BCE and Capricornus at 1357.933 CE. Thus the relation does not provide a second clock 
confirming 2985 BCE.
The complete HYG regional panel also places Cygnus first at the imported epoch and second at J2000. 
Across the six S6 Cygnus state variants, the fixed-epoch mismatch ranges from 0.021921% to 0.118501%, 
while exact-equality epochs vary from approximately 3323 to 5329 BCE. Those six checks change Cygnus 
only, unlike the full field recensus in section 3. No six-variant all-sky ranking is attributed to this axis test.
The Station side ratio is 80.095/33.435=2.395543592, versus ℓ/τ=1.373058775. Reading the two stellar axes 
as rectangle sides misses the ratio by −42.683%; reading them as its two diagonals requires equality and 
misses by +37.306%. A centred rectangle also has four equal centre-to-vertex distances, whereas the stellar 
radii differ by a maximum/minimum factor 2.646. These two explicit proportional templates fail. No flexible 
projection was fitted to rescue them.
In the right-ascension channel, (7+18)/4=6.25 and 7/(7+18)=0.28 merely restate the target’s axis ratio and 
hub position when the adjacent intervals add. They are not the great-circle quantities above and cannot be 
counted as further independent matches. The spherical near equality adds a descriptive property of Cygnus, 
without identifying a new homologous feature in Stonehenge.

Stonehenge | S8 | v1.39 | 7
S8.7 Inverse construction and withheld-channel controls
A further retrospective test inverts the existing continuous construction, with an irrelevant positive common 
scale omitted: (d₁,d₂,d₃)=(2z−y,3y,z−x). The inverse is y=d₂/3, z=(d₁+y)/2 and x=z−d₃. The right-angle 
condition z²=x²+y² is equivalent to 9d₁d₃+3d₂d₃−9d₃²−d₂²=0. Retain only positive dimensions with x<y. These 
are conditions on a hypothetical inverse rectangle, not a new survey measurement.
For every one of the 256 Hipparcos J2000 fields, search all sign-changing roots over the full time window 
with 0.5-year and 1-year grids and bracket refinement. Both grids return 21 valid roots in 14 fields. This 
scope does not certify the absence of tangential roots or crossings missed between grid samples. Compare 
the inferred y/x afterward with the published 2.395543592; neither the aspect nor chronology selects the 
candidates.
Cygnus has a root at 2936.745850 BCE giving y/x=2.551117104, 6.4943% above the published aspect. This 
is the closest of the 21 retained roots, but it is not precise metric recovery and does not improve the earlier 
dimensional fit. Its other root is near 6703.491 BCE with aspect 2.846313370, a discrepancy of 18.817%. 
The complete root table is supplied; the inverse is not used to alter the field census or counted as an 
independent positive result.
Fitting two channels and inspecting the third
An auxiliary Cygnus test equates each pair of normalized intervals and inspects the unused channel. The 
first two channels never reach exactly 7/18 in the sampled full window; their ratio spans approximately 
0.36128–0.38295 against 0.38889. Fitting channels 1 and 3 returns roots near 6498.313 and 2997.929 BCE, 
leaving channel 2 discrepant by +3.7973% and +1.6004%. Fitting channels 2 and 3 returns roots near 
6573.388 and 2973.113 BCE, leaving channel 1 discrepant by −3.7844% and −1.5794%. All roots and 
residuals are retained.
These calculations expose multiple roots and imperfect third-channel agreement. The two fitted ratios reuse 
the same inputs and target, and the control was designed after the full fit was known. It is therefore a 
sensitivity diagnostic rather than prospective validation of a withheld observation.
What the combined audits establish
The dimensional agreement survives a continuous exterior construction. The ratio objective has a 
coordinate-invariant form about the physical pole. A specified brightness-and-field census recovers the five-
star group without regional boundaries: fixed-scale Cygnus rank is 1 in the complete 19,939-group census, 
whereas free-scale rank is 68 there and 1 in the four-arm subsets. Local ratio ranking survives the targeted 
state variants within the cap-centred four-arm family. Five-member star-centred circular fields, the ordered 
count map and geometric roles remain model choices. None of these comparisons establishes prehistoric 
use, transmission or a construction date; ranks are not independent probabilities.

Stonehenge | S8 | v1.39 | 8
S8.8 Reproduction and integration record
Extract S8_Geometry_and_Field_Census_Code_v1_39.zip. The English root README maps four modules. 
Modules 01_Invariance, 02_Axes and 03_Fields preserve their prior code, inputs, outputs, hashes and 
reproduction records byte for byte. Historical Portuguese statements that the manuscript was unchanged 
describe those modules’ original releases. Module 04_FixedScale adds independent complete enumeration, 
fixed/free-scale comparisons and both grids. Keep the four directories together: 04_FixedScale imports the 
archived engine, catalogue and selector from sibling 03_Fields. No external data download is required.
Table S8.7. Run commands from each named module directory
Directory
Commands in order
01_Invariance
python audit_robustness.py
02_Axes
python audit_axes.py
python audit_kinematics.py
03_Fields
python all_sky_caps.py
python kinematic_fields.py
python diagnose_inverse.py
04_FixedScale
python audit_open_fields.py 1
python audit_open_fields.py 0.5
python verify_results.py
The executed environment is Python 3.12.14, NumPy 2.3.5 and SciPy 1.17.0. After dependency installation, 
no network is required. Enumeration uses a roughly 210 MB distance matrix; the complete half-year census 
additionally allocates about 1.6 GB for two angular arrays, plus other working data. Catalogue provenance 
and the HYG licence accompany the inputs.
A fresh 03_Fields run reproduced all 26 result files byte for byte. Modules 01_Invariance and 02_Axes retain 
their earlier execution records. Module 04_FixedScale supplies all 19,939 assignments, both score grids and 
a numerical verifier. Integration checks preserved bytes, tables and rendered documents; no new full 
execution of S1–S7 is claimed.
Main sections 2.9 and 3.16–3.17 report S8. The retained S7 appendix’s v1.37 information-order reference 
2.9 now corresponds to main 2.10. Earlier candidate families keep their stated definitions. Reproduction 
digits are not observational or archaeological accuracy.
INPUT_ARCHIVES.json records source ZIP hashes; SHA256SUMS.txt covers released files. The overall 
v1.39 package records edits and integrity checks. Its input v1.38 ZIP has SHA256 
bad839691d5e19d62b2e38807f517853e2e08a340c27da7ea835bac3091f8610.
No original georeferenced Station or Aubrey survey was obtained. These calculations use existing 
measurements and astrometry and require no further archive response.

Stonehenge | S8 | v1.39 | 9
S8.9 Complete fixed angular census and matched free-scale control
This extension was executed after the four-arm ratio findings were known. It uses the same archived 
nominal Hipparcos V≤6 catalogue, angular proper-motion model, precession engine and geometric selector. 
Independent cap enumeration returns 19,939 distinct quintets across 25,161 generating radius intervals. 
The centre remains among the five brightest entries in each generating field. This is the complete family 
under that definition, not all five-star subsets of the catalogue.
At J2000, each quintet receives one assignment: unique four-arm minimum spanning tree hub if present, 
otherwise angular medoid; brightest non-hub endpoint; most opposite tangent-bearing endpoint; remaining 
unordered pair. The last pair is stored in increasing HIP order without changing its unsigned interval. Of the 
19,939 groups, 283 have a unique four-arm hub at any member; 256 additionally have that hub among their 
generating cap centres. The latter assignments exactly match the archived 03_Fields panel after 
canonicalizing the residual pair.
For absolute wrapped right-ascension widths d(t) in degrees and θ=(7°,18°,4°), minimize 
R(t)=√{Σᵢ[dᵢ(t)−θᵢ]²/3}. Here R denotes angular RMS; the R columns in Tables S8.1–S8.2 denote error ratios. 
Then 4R is the numerical RMS in quarter-degree units, equivalent to the original sidereal-minute 
representation of W=(28,72,16). Conversion of units is exact and leaves ranks and epochs unchanged; 
choosing an absolute target is still a model assumption. The matched free-scale search minimizes the same 
E defined in section 2 over exactly the same role assignments.
Both objectives are searched over astronomical years −8000 to +2000 on 1-year and 0.5-year grids, refining 
every bracketed minimum and endpoints with the archived optimizer. The fixed-scale horizon branch locates 
and refines all intervals in which every member has positive geometric upper culmination at latitude 51.18° 
N, then reoptimizes R within them, including boundary infima. Neither catalogue grouping nor optimization 
uses archaeological chronology. Groups and roles are held fixed during their temporal searches.
Table S8.8. Nominal Hipparcos J2000 complete census and subsets. Error ratio is runner-up/Cygnus 
under the stated objective; a dash indicates that Cygnus is not first. Only the explicitly labelled row 
imposes the horizon condition.
Objective and candidate family
Groups
Cygnus rank
Error ratio
Fixed target; all fields
19,939
1
2.201
Fixed target; horizon reoptimized
17,860
1
2.626
Fixed target; four arms, any hub
283
1
12.915
Fixed target; four arms, cap-centred hub
256
1
12.915
Free scale; all fields
19,939
68
—
Free scale; four arms, any hub
283
1
1.0155
Free scale; four arms, cap-centred hub
256
1
1.0155
Cygnus has fixed R=0.09753151601066437° (0.3901260640426575 quarter-degree units) at astronomical 
year −2964.4251427554113, or 2965.425143 BCE. The unrestricted runner-up is HIP 
17440;16368;15968;13884;16245 in role order, spanning Reticulum, Hydrus and Horologium. Its 
R=0.21469519640043608° at 2430.125639 BCE gives ratio 2.201290467. Modern regional labels are 
descriptive only.

Stonehenge | S8 | v1.39 | 10
The fixed-scale horizon branch contains 17,860 groups with at least one admissible epoch. Cygnus remains 
first; HIP 56211;52425;47654;50933;52353 is second at 1534.776358 BCE, with R=0.25612245061204775° 
and ratio 2.626048083. In contrast, 13,910 groups satisfy the condition at their unconstrained fixed optimum. 
Filtering only those optima would omit groups with admissible alternatives, so it is not the implemented 
horizon comparison. The necessary horizon test does not establish simultaneous night-time observability.
Both four-arm subsets have fixed-scale runner-up HIP 85670;87833;94779;86414;87585, with 
R=1.2596245287598156° at 1833.144791 CE and ratio 12.915051260. These subsets are nested in the 
complete nominal census and do not supply independent confirmations.
The free-scale Cygnus optimum is E=0.006490264894093756 (0.6490264894%) at 2985.482488 BCE. 
Cygnus ranks 68th among all 19,939 groups and first in both four-arm subsets, whose runner-up is centred 
on HIP 109268 with ratio 1.015506711. The free and fixed Cygnus optima differ by 20.057345 years and 
both fall in 3000–2920 BCE. Epoch compatibility under these two objectives does not imply scale-
independent identification, an uncertainty interval or independent archaeological dating.
The 1-year and 0.5-year searches recover all stated ranks, denominators and leading error ratios. Across all 
groups, the largest fixed-scale score difference is 1.7582×10 ⁸ degree, also the maximum for horizon 
⁻
reoptimization; the largest E difference is 6.7315×10 ¹². The score tables retain every candidate. Numerical 
⁻
grid agreement is not a proof excluding unsampled continuous-time extrema, and reproduction digits do not 
represent archaeological or astrometric accuracy.
This complete census has been executed for nominal Hipparcos/J2000 only. The four catalogue/geometry 
panels and six kinematic variants in sections 2–3 test the earlier cap-centred four-arm ratio family. Their 
robustness must not be attributed to the full fixed-scale ranking. Alternative target orders and geometric-axis 
controls in sections 4–7 also retain their own stated universes.
The role rule is explicitly documented in section 2.9 (Nested references and joint persistence) of the 
separate stellar manuscript identified as main reference 47; section 2.10 uses the resulting trio and pair. 
This source verification supports attribution of the algorithm. It does not make the present retrospective field 
census a preregistered or blinded archaeological test.
