Compose source-clear Basic even-force Ecaz Occupy

This commit is contained in:
Quintine committed 2026-10-03 20:34:22 +10:00
1 parent 63e57bd894
commit 5f082696a6
18 files changed
+861 -52

No files matched your search

+47 -4
View File
@@ -1,6 +1,49 @@
# Ecaz Occupy: Advanced source and historical conflict
# Ecaz Occupy: Basic and Advanced source, and historical conflict
Historical primary-source audit, 2026-09-07, corrected after independent verification. The old publisher/designer casualty-rounding conflict remains unresolved as a publisher correction and for Basic. Its earlier false attribution to Jack Reda is still withdrawn. The later user-authorized Advanced source below selects explicit Advanced arithmetic; it does not retroactively certify that attribution.
Historical primary-source audit, 2026-09-07, corrected after independent verification. The old publisher/designer casualty-rounding conflict remains unresolved as a publisher correction and for odd Basic counts. Its earlier false attribution to Jack Reda is still withdrawn. The later user-authorized Advanced source below selects explicit Advanced arithmetic, and the Basic even-force composition selects E/2 for even counts; neither retroactively certifies that attribution.
## Basic even-force composition — 4 October 2026
Fresh local `ecaz-occupy` now admits Basic as well as Advanced. Basic uses the
exact `ecaz` Treachery deck with classic or optional Moritani opponents and no
optional cards or modules. Genuine setup, original Nexus/worm alliance,
shipments, Karama and Collection remain native; older played Basic tables are
not retrofitted.
For an even Ecaz force count the publisher page-8 general paragraph and the
page-15 example agree: contribution and ordinary-win Ecaz losses are E/2, and
E/2 survive. The two pages still contradict each other for odd counts, so an
uncanceled odd Basic count is rejected at the original battle choice before a
lead frame is saved. It is not silently rounded, canceled or replaced with
another fight. Advanced keeps its separately authorized ceiling arithmetic.
Basic has no Emperor/Fremen elite inventory (elites require Advanced or
Homeworlds), and the engine normalizes a supplied Basic support amount to zero
rather than inventing paid Basic support. The shared editor therefore blocks
nonzero support; no additional Basic payment rule is created. Native Karama
after the lead choice, selected-lead own-pool cancellation, original winner
cards, traitor/mutual/explosion outcomes, joint three-stronghold coexistence and
original Mentat victory all use the existing handlers.
Focused source-band cases plus native runtime cases cover E2/E4 under both
leads, free full-strength ordinary commitments, typed Basic Fedaykin/Sardaukar
strength-one casualties, cancellation under both leads, and the four legal
policies. Five genuine Basic two-through-six-seat games were run: four complete
with 2,090 accepted actions, no rejections and 54 JSON continuations; the
four-seat game is captured at the preserved odd-count conflict and is not rerun
or bypassed. Four completed natural games contain no coalition lead choice, so
the actual coalition arithmetic evidence comes from the controlled programs and
the human path rather than those games.
Authenticated human390px **FEJ6SBFB v14** kept the admitted Ecaz/Guild/Emperor
seats and native wallets, then staged a conserved even-force board after the
genuine Nexus alliance. The table showed the correct Basic labels: a total-dial
slider bounded 2..6 (2 fixed Ecaz strength for four Ecaz forces plus Guild's
four-force variable pool) with no spice-support control. Real sealed plans —
Ecaz Sanya Ecaz 4 with total dial 4 against Emperor Hasimir Fenring 6 — produced
an 8–6 Ecaz win. The log and saved state record Ecaz's two mandatory Occupy
casualties, Guild's two variable casualties, and the Emperor's eight territory
forces to the Tanks; every seat remains at 20 physical forces after refresh.
## Authorized Advanced cutover — 2 October 2026
@@ -232,7 +275,7 @@ Ecaz losses on ordinary allied win = ceil(E / 2)
Ecaz survivors on ordinary allied win = floor(E / 2)
```
The historical publisher/mirror example still conflicts for E=5, and the earlier audits correctly record that publisher gap. Basic still needs a supported interpretation. Advanced now uses the separately user-authorized source cutover above: its explicit ceil contribution and floor survivors select the arithmetic without claiming a designer correction. Fixed Ecaz losses stay independent of the ally's variable dial losses.
The historical publisher/mirror example still conflicts for E=5, and the earlier audits correctly record that publisher gap. Basic now has the bounded even-force composition above: E/2 is source-clear for even counts, and the still-contradictory odd counts are rejected rather than rounded. Advanced uses the separately user-authorized source cutover above: its explicit ceil contribution and floor survivors select the arithmetic without claiming a designer correction. Fixed Ecaz losses stay independent of the ally's variable dial losses.
## Occupation, controller and battle order
@@ -285,4 +328,4 @@ Implement a pure, owner-labelled Occupy battle profile shared by validation, AI,
Acceptance examples should include Ecaz counts 1–5, either selected faction, all three relative storm positions, both win/loss outcomes, Basic/Advanced supported dials, Fremen free strength, advisor-only BG exclusion, printed and BG-converted cancellation, and JSON recovery after selection. Separate targeted follow-ups should cover the source-scoped special battle effects above, agreed/default desert allocation, Advanced Collection, and actual three-co-occupied victory. Keep the optional Homeworld/Nexus/Discovery modifiers explicitly outside the initial profile until their own faces and outcomes are wired; no ordinary Occupy rule may be used to bypass them.
The September audit changed no runtime and corrected the false designer attribution. The October Advanced cutover above is separate; Basic, optional/exotic interactions and independently unanswered cancellation details retain their historical boundaries.
The September audit changed no runtime and corrected the false designer attribution. The October Advanced cutover and the October Basic even-force composition above are separate; optional/exotic interactions and independently unanswered cancellation details retain their historical boundaries.