# Ecaz joint stronghold victory and public progress

Implemented and reviewed 7 September 2026 within the gated Ecaz development scope. The full game goal remains active. This slice adds the three jointly occupied stronghold win; it does not complete Ecaz, its Fremen special-endgame exception, Advanced rules or expansion/module acceptance.

## Source and behavior

[Source contract](ECAZ_VICTORY_RULES.md) records GF9 E3 pp.8,15–16 and applicable base/E1/E2 rules. Both Ecaz and its current reciprocal ally must have effective non-advisor presence in each of three distinct uncontested strongholds. Their ordinary four-stronghold union remains available. One faction holding all three Tech Tokens supplies one ordinary stronghold equivalent; split ownership does not, and tokens never supply a jointly occupied territory. A concealed No-Field supplies public presence without revealing its denomination. Only a placed Hidden Mobile Stronghold counts. Existing storm/control, advisor release and private BG prediction precedence are retained.

`game/victory-progress.ts` is the shared pure public calculation. It returns ordered member IDs, uncontested stronghold IDs, the ordinary threshold, complete-set bonus, jointly occupied IDs, the Ecaz threshold and prospective eligibility. It validates the entire faction/alliance roster, public board, advisor shapes/locations and token ownership. It reads no predictions, hands, traitors, resources, concealed token inventories or token spice. Arrays are detached.

`game/victory-quote.ts` uses these qualifiers before the existing private BG interception and distinct final-turn fallbacks. The existing commit path awards the result once and settles final Stronghold Cards from the same board. An actual Ecaz joint win adds a chronicle explanation naming its strongholds. A BG replacement win does not falsely receive the Ecaz explanation.

`viewGame` projects public progress from the original board, before local advisor normalization; malformed public state produces no unreliable preview while authoritative victory rejects it. `components/victory-progress.tsx` displays one row per side, separates ordinary and joint counts, and explicitly describes a prospective board calculation. It awards no early win. Its native disclosure supports keyboard controls and responsive layout. The internal victory and faction/movement guides describe the supported boundary.

AI destination threats, Trip to Gamont target assessment and Guild timing now consume the same public progress. Hard/Brutal additionally value joining their Ecaz ally in an incompletely shared stronghold. These scores grant no movement permission or hidden information; legal entry and shipment controls still apply. This is a verified policy behavior, not an AI strength calibration claim.

## Verification

- `tests/victory-progress.test.ts`: 11 tests of thresholds, ordinary alliances/player counts, union versus joint occupation, tech ownership, storm survivors, advisor releases, public marker presence, placed HMS, malformed state, private-field getters and detached outputs.
- `tests/ecaz-victory.test.ts`: 17 tests with actual Basic/Advanced Collection-to-Mentat actions, prediction replacement, force preservation, No-Field 0/3/5 and final cards, printed and BG-converted Moritani placement cancellation, Hard/Brutal third-joint shipments and real Guild timing.
- `tests/ecaz-victory-recovery.test.ts`: 3 production room/authentication/SQL tests in memory. Includes Basic/Advanced races with exactly one successful CAS and winner log, stale/current duplicate rejection with zero writes, final card ownership, card/force conservation, restarted module/JSON restoration, unrelated room preservation and identical complete rival views under private prediction changes.
- Full registered suite: **2,758 rules/client/component +245 multiplayer =3,003 passing tests**. TypeScript, lint and production build pass. Logs: `/tmp/dune-ecaz-victory-{rules,multiplayer,type,lint,build}-final.log`. Rules 80.698 seconds; multiplayer 25.559 seconds.
- Independent integration review: `/tmp/dune-ecaz-victory-integration-review.md`. It found a malformed-advisor preview mismatch, corrected by strict stance validation and original-state projection. Exact reproductions now reject/suppress the preview. No further concrete privacy, winner-order or post-cost defect was reproduced. Root reviewed all new helpers and tests.

The browser used only the existing named QA room. Its prior contested Carthag correctly showed zero qualifying holdings. A backed-up, conserved three-joint position was staged at Collection, version 116; the real final readiness action produced version 117, the Ecaz/Emperor win and exactly one explanatory chronicle entry. Phone 390×844 layout, keyboard Space disclosure and real refresh passed; the viewport was reset. All **2,526 pre-existing non-QA** room versions/state hashes remain unchanged; the regression suite created 30 additional rooms. No human room was modified. Restricted artifacts are `/tmp/dune-victory-qa-{before,staged,final}.json` and `/tmp/dune-victory-other-rooms.json`.

Twenty balanced Basic full games (seed 20261023, two through six players, all four homogeneous difficulties) completed **12,818 accepted actions, zero rejected candidates/stalls/exceptions/invariant failures and 568 JSON round trips** in 120.643 seconds. The source fingerprint `5b63c390f73da1ef0f66239c87f6a53acce34f422f1fc9cd3380ba3724c0f788` was unchanged during the run and matched the root’s post-run file check. Artifacts: `/tmp/dune-ecaz-victory-fullgames-final.{json,log}`; runner `/tmp/dune-sapho-base-fullgames.ts`. These are base-game regression results; they do not demonstrate full Ecaz/Advanced/expansion play or comparative AI strength.

## Remaining boundaries

The Fremen special-endgame exception remains unimplemented in this slice. The user has a pending interpretation question about co-occupation in Habbanya and Ecaz-only occupation; elapsed time is not an answer. The explicit source permission for co-occupied Tabr is recorded separately. Combined Occupy combat still needs a participant model and resolution of the documented numerical source conflict; shared desert collection, Tleilaxu Ambassador integration and broader faction/module interactions remain unfinished. No public mode gate or deployment changed. The removed automation remains removed.