Skip to content

A Civilization-style strategy game about governing Israel from 15 May 1948. Real ministers, real map, every fact sourced. Unity + deterministic C# sim.

108 KB
Raw Blame

Determinism

The simulation must produce the same state, bit for bit, on .NET, Unity Mono and IL2CPP. Saves, replays and the "Why did this happen?" report depend on this.

How it is kept

  • One RNG: SplitMix64 (Statecraft.Sim.Rng). Its whole state is one ulong in GameState.RngState.
  • Reference vectors: DeterminismTests.SplitMix64MatchesTheReferenceVectors checks the first three outputs for seeds 0 and 1234567. The expected values come from Sebastiano Vigna's reference splitmix64.c (https://prng.di.unimi.it/splitmix64.c), compiled and run separately. They were not copied from this implementation.
  • No trigonometry in the sim: paths that change state call only Math.Abs, Math.Floor, Math.Max, Math.Min and Math.Round, and every such Math.Round passes MidpointRounding.AwayFromZero. PlayerText also rounds, for display only, and MapOverrides uses Math.Sqrt(3) only to diagnose editor offsets; neither changes state. Geometry that needs cos/sin (which hex owns a settlement, which hex a starting force's anchor is on) is precomputed by tools/settlement_hexes.py and tools/unit_anchor_hexes.py and read as content.
  • Canonical hash: GameState.Hash() is SHA-256 over a canonical text. Keys are sorted ordinally. Doubles are written as IEEE bit patterns, with -0.0 written as 0.0. Culture is invariant. Fractional site and city-project carry follows the same rule: nonzero carry is written as 16 hexadecimal digits of its IEEE bit pattern, and zero carry is omitted. Round-trip decimal formatting is only a display concern: .NET and Unity Mono can spell the same carry bits differently. The Phase 7 minister carry correction re-pins hash1950, hash1952 and mvp-end for this representation only; stored values, gameplay and completion dates are unchanged.
  • Settlement control: the player capture check runs after the route step and each order, then again in resolver step 3 after enemy combat. Each check visits targets in ordinal id order and reads only unit positions, with no RNG draw (R-CTRL-004; CaptureTests.SettlementsResolveInOrdinalOrder, NoRngDraw and TheInstantCheckDrawsNothing). The shipped list has six targets (#793), and only mvp-occupy-cost and mvp-annex-nazareth capture a place (Nazareth, whose residents stayed, so no golden records a flight, fled=1, #799): player units may enter the five enemy-held ones since #1304 (R-MOVE-005), but the reference player sends no Move, so none of its units stands on one, and its unit on Jaffa's hex (shared with Tel Aviv) captures nothing, since Jaffa is not enemy-held at the start. The smoke-orders move skips the lit hexes of enemy-held places on purpose (GoldenRun.SmokeMove), so it stays a plain move. The enemy's check (R-CTRL-006, #1305) runs in the same step after it, visits the player's localities and the annexed targets (the owner on #2955, "Can be taken") in one ordinal id order and draws nothing (CaptureTests.PlayersLocalitiesResolveInOrdinalOrder, ALossAndARetakeDrawNothing, AnEnemyUnitAloneOnAnAnnexedTargetTakesIt). An enemy move ends on a player's locality or an annexed target only where it is sent, and no scripted anchor is on one, so no golden records a loss and no hash moved, mvp-annex-nazareth's annexed Nazareth included (CaptureTests.RealContentLosesNoPlayersLocalityOnTheFirstTurn).
  • Answering a capture (Raid and withdraw): the units on the place leave in ascending id, each to the first free hex in the fixed neighbour order, and the captures still waiting at the game's end resolve in ordinal id order, with no RNG draw (R-CTRL-008, R-CTRL-010; RaidTests.RaidDrawsNothing). Only mvp-occupy-cost captures a place, and it answers with Occupy, so no golden Raids one.
  • Occupation timers (resolver step 10): after step 9's arrivals and before objectives and score, each occupied place in ordinal id order pays its upkeep and counts its turns= down, or ends as a Raid when no Israeli military unit holds its hex, with no RNG draw (R-CTRL-011 to R-CTRL-013; OccupyTests.TheTimerCountsDownInStepTen and OccupationDrawsNothing). mvp-occupy-cost (#798) is the golden that occupies a place: it plays the real content, moves Golani's unit onto Nazareth, answers the capture with Occupy and runs to the end of the timer; OccupyTests plays the same run on fixture content. mvp-annex-nazareth (#799) plays the same run and annexes Nazareth when the timer ends (R-CTRL-016): one new city from the row's count, with no RNG draw (EndOccupationTests), and the same count joins GameState.Population's annexed group (R-CITY-008, #2608), which moved this golden's population, debt, FX and Food lines. No other golden occupies a place, so none moves.
  • Deport (resolver step 10, after the occupation timers): the units leave as in Raid; the costs, the drag (each turn, per place in DisplacedByPlayer) and the International Support ceiling are plain arithmetic on the indicator's shift, with no RNG draw (R-CTRL-018; DeportTests.DeportDrawsNothing). DisplacedByPlayer is hashed as displaced= only when non-zero. The real content keeps Deport out of play (mvp.deport_in_play: false), and no golden Deports, so none moves. The mvp-deport-cost fixture comparison (not a manifest golden), paired with mvp-occupy-cost, cannot play the real content, where Deport is refused (#800); DeportNeverDominatesTests plays both runs from the same fixture states.

Combat draws

A player engagement draws Combat.DrawsPerEngagement (1) when its order runs in RunOrders, after the route step and any earlier committed order, before every waiting order. An operation card's attack (R-ENEMY-008) draws the same way when the card is decided: in its option's list order when its choice runs in RunOrders, or, when the card is delegated, after that delegation's own draws (EventSystem.Delegate hands the stream to the attack and keeps its draw). The enemy's engagements draw in resolver step 2, in ascending attacker id.

Golden playthroughs

Statecraft.Sim.GoldenRun.Play(content, until) starts Simulation.NewGame(content, GoldenRun.Seed, Mode.Historical) with GoldenRun.Seed = 1948 on the real content/, then steps with GoldenRun.Orders until Date >= until (or the game ends, or 1000 turns pass). Since the GDD cutover (Phase 2 PR 1) the orders are the reference player's (Statecraft.Sim.Reference.ReferencePlayer, decision D-A7): a stated test policy, not game AI and not history. It sends HistoricalAI's orders and leaves the jobs to each city's default (R-CITY-009), then trains one halutz group when there is none and none in training, while the turn's Attention lasts (training borrows, so the Treasury does not hold it back). No game mode or UI reaches it (ReferencePlayerTests); every copy of the golden loop (the writer preview, the invariants, the era goldens) takes its orders from GoldenRun.Orders. The real game plays the GDD loop on monthly turns from 15 May 1948 and ends on 15 January 1952 after 44 turns (R-LOOP-004). Each run's final GameState.Hash() is pinned:

Run Ends Hash
to 1950-01-01 (Hash1950) 1950-01-15, turn 20 832e8ccc38e3fb36228826816b452477fbbd6ab667ee30cdaef40e5b1b2fe869
to 1952-01-01 (Hash1952) 1952-01-15, turn 44 (the game's end) 841f05536a49b68cd10d3f1ca548e559098415119284d483620f14bb6a4bb44e
to the game's end (mvp-end) 1952-01-15, turn 44 841f05536a49b68cd10d3f1ca548e559098415119284d483620f14bb6a4bb44e

hash1952 stops at the first turn on or after 1 January 1952, which is the game's last turn, so it is the same run as mvp-end (DeterminismTests.TheWholeGameToItsEndIsFixed). Both names are kept (spec 2026-10-12 section 3 step 6): a later end date moves only mvp-end. mvp-end is pinned in the manifest and its snapshot; Unity EditMode and every player check enumerate the manifest, so it and every scenario are checked without another hand-written list.

A further value is pinned for the browser smoke run (?smoke=5: the golden seed, five turns played through the UI's own buttons, answering each event with its first option while the turn's Attention pays for it and that option's operation units are in place in the state the card shows: the turn's start after its standing go-to routes have moved, the no-orders preview (R-ENEMY-007, GoldenRun.SmokeOrders; the smoke gives no unit orders), the rest left to their ministers): SMOKE OK 5 turns 1948-10-15 12f088f62bd0694ae59b5f6472751579ad96c2198308d0869a8cf795f1aad051. It depends on the UI path as well as the sim (which option is first, the turn grid), so re-record it when either changes. Check every WebGL build with every manifest entry (plus the UI-path smoke line) with:

golden_expect=()
while read -r name hash; do
  [[ -z "$name" || "$name" == \#* ]] && continue
  golden_expect+=(--expect "GOLDEN OK $name $hash")
done < sim/Statecraft.Sim.Tests/golden/manifest.txt
uv run tools/web_check.py unity/Builds/Web-<compression> --query "golden&smoke=5" "${golden_expect[@]}" \
  --expect "SMOKE OK 5 turns 1948-10-15 12f088f62bd0694ae59b5f6472751579ad96c2198308d0869a8cf795f1aad051"

smoke-orders (GDD Phase 5 PR 15, #754) is the same kind of browser run for the unit orders (?smoke-orders, desktop -smoke-orders: the golden seed, one turn). It hides the turn's event card (the events go to their ministers), clicks the hex of the unit GoldenRun.SmokeMove names (the lowest-id unit Movement.Reachable has a hex for: unit 0, Golani, on 11,11, its first anchor in content/map/unit_anchor_hexes.yaml), presses the unit card's Move button, clicks the first lit hex in Reachable's (q, r) order that is no enemy-held place (8,12, no place, so the run captures nothing) and presses End Turn: SMOKE ORDERS OK 1948-06-15 08d9800eda4462ef165cc7ea6285fad059c030c0f87ad563e35c9789cc352d78. GoldenRun.PlaySmokeOrders plays the same turn with that one Move, so the value is in the manifest (SliceSmokeTests.TheSmokeOrdersRunMovesOneUnitAndEndsOnTheGolden) and the player's golden check prints it as GOLDEN OK smoke-orders. Re-record it when the sim or the order path on the card changes; the fortify, hold and attack slices can add their orders to it. Check a WebGL build with:

uv run tools/web_check.py unity/Builds/Web-<compression> --query "smoke-orders" \
  --expect "SMOKE ORDERS OK 1948-06-15 08d9800eda4462ef165cc7ea6285fad059c030c0f87ad563e35c9789cc352d78"

When a golden moves. The test fails with the lines that changed, grouped by section ([ind], [pol], [bud], [sit] and so on), against a committed snapshot of the run's final state: sim/Statecraft.Sim.Tests/golden/hash1950.txt and hash1952.txt, written by GameState.Describe(), which uses the same writer as Hash() (so the two cannot list different fields) and shows every number as a plain value next to its IEEE bits. If the change is deliberate, run STATECRAFT_UPDATE_GOLDEN=1 dotnet test sim/Statecraft.sln: it rewrites the snapshots, so the PR's diff of them shows reviewers what moved. The same command rewrites sim/Statecraft.Sim.Tests/golden/manifest.txt (the pinned hashes) and the GoldenRun.cs constants. That run still fails for every golden that moved, against the value it started with; the tests never take a hash from the run without the variable, so a second plain run is the check. Then update the docs below by hand (GoldenManifestTests names any place still showing the old hash), add a row to the rebaseline history, and say why in the commit message. The snapshot test also fails if a snapshot is stale while the hash is unchanged.

Adding an era golden. In a test, call EraGolden.Check("hash_<era>_<year>", content, "<era id>", new DateTime(<year>, 1, 1)). It plays a fresh start of that era (golden seed unless one is passed, Historical mode, GoldenRun.Orders, the same loop as GoldenRun.Play) and checks the hash against the manifest line of that name and the snapshot golden/<name>.txt. The run stays inside the era: it stops before the turn that would reach the era's to, because that turn ends on the boundary and its state already holds the next era's attention and events, so the next era's rules would move this era's golden (EraGoldenTests.TheRunStaysInsideTheEraSoTheNextErasRulesDoNotMoveIt). Bless it with STATECRAFT_UPDATE_GOLDEN=1 dotnet test sim/Statecraft.sln, which adds the manifest line and the snapshot; list the name in GoldenManifestTests.TheManifestHoldsExactlyTheKnownGoldens. Today only the first era can be started fresh, because NewGame reads the base rules' start; a later era fails with a message saying its start values are not read yet. The carry-over check (play era N into N+1 and compare with a fresh N+1 on each carried quantity, chapters design section 6) waits on that and on EraSystem.Cross.

Scenario goldens. The hash goldens follow the reference player, HistoricalAI's orders on the path history took. A scenario is a path a player might take instead: sim/Statecraft.Sim.Tests/scenarios/<name>.yaml lists commands by date (on:), the runner (Scenario, in the sim library) plays them from a seed (default 1948) in Historical mode until until:, and ScenarioTests checks the final hash against the manifest line scenario-<name> and the snapshot golden/scenario-<name>.txt. The file's header comment lists the command kinds. A turn with no scripted command passes. A command the sim rejects fails the test with the date, the order and the reason, and a step that never runs fails too. To add one: write the file, run STATECRAFT_UPDATE_GOLDEN=1 dotnet test sim/Statecraft.sln --filter ScenarioTests (it adds the manifest line and the snapshot), then run the tests plain. A scenario that moves is re-pinned with the hash goldens by the same command; its snapshot diff in the PR shows what moved, and the commit message says why. All six play the real content. cautious and reckless run through March 1949, setting rationing on the first turn after its office opens; mvp-absorb-first (GDD Phase 1 PR 3) sets Tel Aviv's workers to 90 of its 106 job slots on the first turn (all are filled at the start), freeing 158 civilians for that turn's halutz training, and completes the absorb objective on 15 January 1950; mvp-defeat (GDD Phase 7 PR 5) overspends and turns down every source of foreign currency until Money and Food run out, and ends in defeat_economy (R-END-002).

MVP goldens after the cutover. The real game is the GDD game, so every golden above is an MVP golden. Each PR that changes an MVP rule or a design value (R-LOOP-005) moves the goldens it reaches, re-pins them in the same PR, adds a row to the history below and says why in the commit and the PR (spec 2026-10-12 section 3, "After cutover"). PRs merge one at a time, so there is no "re-record at integration".

Before cutover. The last legacy goldens (the legacy loop, HistoricalAI alone, weekly, monthly and quarterly turns by security level, play to 1957) were hash1950 71e9904d0ee48083db9bbba281f3f046136482ba83ee056e48d57b62e76a5b07 (1950-02-27, turn 35), hash1952 62f8ec61a14acff4eab503381994a7c94d7a2f786dcfde5c4c19439ac8b4c5e8 (1952-02-27, turn 43), smoke5 SMOKE OK 5 turns 1948-06-19 d3fdd459ddb876bd8d775c20edd1202cf56ca7134a8ad06f873d4d92034aa5b6, scenario-cautious 1d5b8d71, scenario-reckless c0bd12bc, and on the Phase 1 overlay mvp-slice ec6f8c80 (after #566; retired with the overlay, R-LOOP-003) and scenario-mvp-absorb-first 59ea3181. The owner tags the last pre-cutover commit pre-gdd-v0.1.

Scenario lives in Statecraft.Sim next to GoldenRun, so Unity, which gets only Statecraft.Sim.dll and YamlDotNet.dll from tools/build_sim_for_unity.sh, can play the same runs (#394). ContentSync copies the manifest and the scenario files into the player (Assets/StreamingAssets/golden) before every player build, for the player's own golden check (?golden on WebGL, -golden on desktop). Of these, only the snapshots (the golden/*.txt files other than the manifest) stay with the tests.

Invariants. A hash says something changed; sim/Statecraft.Sim.Tests/golden/invariants.yaml says whether it is still plausible. Each entry names a golden run (1950 or 1952), a value (an indicator id, or treasury, debt, fx, population, stability), when it is read (end, always, or a date: the first state on or after it), a min and/or max, and a reason (a source, or the test or note the band comes from). InvariantTests plays the golden runs and fails with the check, the value, the date and the band, for example invariant 'foreign-currency-runs-out' (...): fx = 12.5 on 1952-02-27 is outside [-inf, 0]. The invariants only add checks: a tuned rate that stays in its band still moves the hash goldens (rebaseline as above), but not an invariant. The shipped bands are loose guards set around the current runs; the design document and the maintainers decide which ranges matter.

Rules per era. A system reads Content.RulesAt(date), where date is the date of the state the turn starts from (s.Date before the step); AttentionSystem.Refill runs after the date moves, so it fills the next turn from that turn's era. Rules change only at era boundaries (chapters spec section 3.3). Spec section 4.1 ends the turn that reaches a boundary exactly on it: Simulation.TurnEnd cuts each turn at the next era's start or the game's end, whichever comes first, so every turn lies in one era (sim/Statecraft.Sim/Simulation.cs). Start values (start, treasury_start, fx_start, start_rung, initial_population, budget_start, budget_owners) are read once by NewGame from the base rules.yaml: an era override cannot set them. A map action's display name is also read from the base rules, because an override cannot remove an action. EraRulesSwitchTests.OnlyStartValuesAndMapActionNamesReadTheBaseRulesDirectly fails if a file in sim/Statecraft.Sim reads Rules directly for anything else, including a bare var r = c.Rules;.

Cause text is derived, not hashed. The "Why did this happen?" lines (Cause.Message, Cause.Delta) are output computed from the hashed state; replaying a save's commands regenerates them. Wording changes therefore do not move the goldens (playtest 2 changed every line's casing with no hash change). They are still deterministic: FinalWaveTests.AReloadedSaveReproducesTheLastTurnsCauses checks a reloaded save gives the same last-turn Causes, and TheGoldenRunsCausesAreTheSameInEveryCulture runs the 1952 golden under de-DE and ar-SA against the invariant culture.

Each golden hash is pinned in sim/Statecraft.Sim.Tests/golden/manifest.txt (hash1950, hash1952, mvp-end, smoke5, smoke-orders, and the scenario lines above). The .NET tests compare against it. These places must agree with it, and GoldenManifestTests fails naming the one that does not:

  • sim/Statecraft.Sim/GoldenRun.cs, which Unity code and players compare against (the update command rewrites it);
  • this file, docs/RELEASING.md and tools/web_check.py's usage example (by hand).

tools/art_checkpoint.sh reads its expected hashes from the manifest, so it needs no edit.

unity/Assets/Tests/EditMode/GoldenCheckTests.cs reads the manifest and runs every entry inside Unity Mono. GoldenCheck does the same in macOS and WebGL IL2CPP; ContentSync ships the manifest and scenario YAML, and the WebGL loader fetches them before the check. tools/player_golden_check.py and tools/release_build.py also derive their required OK lines from the complete manifest. Run uv run tools/build_sim_for_unity.py first so Unity has the current DLLs.

Verified on .NET, Unity editor Mono, a macOS IL2CPP player (Apple clang, no -ffp-contract=off needed) and a WebGL IL2CPP player in Chrome (M3 Task 22). The platform checks below used the prior 73590e32 1952 baseline. The issue #2 content correction has since been rechecked on .NET, in the Unity EditMode tests and in WebGL (see below and docs/visual-v1.md), and on a macOS IL2CPP player (2026-10-14, below).

Re-checked at the M3 merge candidate (2026-10-11, commit a099655):

  • macOS IL2CPP player (Build.Mac, then -batchmode -nographics -golden): exit 0, printing GOLDEN OK 1950 71e9904d0ee48083db9bbba281f3f046136482ba83ee056e48d57b62e76a5b07 and GOLDEN OK 1952 73590e324f19e9959a455a4da79e93d7089dfcf283c13aac02af6b0a624b9767.
  • WebGL IL2CPP player (Brotli release build, web_check.py with the three full values above): PASS.

Re-checked on macOS after the issue #2 correction (2026-10-14, commit bd683ee0 on the #542 branch, Unity 6000.3.25f1, a --dev build through tools/release_build.py --platform mac, which runs the player with -batchmode -nographics -golden): exit 0, printing GOLDEN OK 1950 71e9904d0ee48083db9bbba281f3f046136482ba83ee056e48d57b62e76a5b07 and GOLDEN OK 1952 62f8ec61a14acff4eab503381994a7c94d7a2f786dcfde5c4c19439ac8b4c5e8 on an IL2CPP Universal player (x86_64 and arm64; this Mac ran the arm64 slice). The run is recorded in #542. These are the legacy values (see "Before cutover"); the GDD values are not yet re-checked on a desktop player.

Re-check with:

uv run tools/machine_lock.py --shared --wait-unity -- "$UNITY" -batchmode -CI false -GITHUB_ACTIONS false -TF_BUILD false -projectPath unity -runTests -testPlatform EditMode -testFilter GoldenCheckTests -testResults "$PWD/unity-editmode.xml" -logFile "$PWD/unity-editmode.log"
uv run tools/machine_lock.py --shared --wait-unity -- "$UNITY" -batchmode -CI false -GITHUB_ACTIONS false -TF_BUILD false -quit -projectPath unity -buildTarget WebGL -executeMethod Statecraft.Game.EditorTools.Build.Web -compression disabled -logFile unity-build-web.log
golden_expect=(); while read -r name hash; do [[ -z "$name" || "$name" == \#* ]] || golden_expect+=(--expect "GOLDEN OK $name $hash"); done < sim/Statecraft.Sim.Tests/golden/manifest.txt
uv run tools/web_check.py unity/Builds/Web-disabled --query golden "${golden_expect[@]}"
uv run tools/machine_lock.py --shared --wait-unity -- "$UNITY" -batchmode -CI false -GITHUB_ACTIONS false -TF_BUILD false -quit -projectPath unity -buildTarget OSXUniversal -executeMethod Statecraft.Game.EditorTools.Build.Mac -logFile unity-build-mac.log
uv run tools/machine_lock.py --shared -- "unity/Builds/Mac/Statecraft.app/Contents/MacOS/Statecraft Israel" -batchmode -nographics -golden -logFile "$PWD/unity-golden-mac.log"; echo exit=$?

Last full-set run (all seven manifest goldens, one GOLDEN OK <name> <hash> line each, every hash equal to the manifest), 18 October 2026, commit 6d29976b (#394) on 30fbe580: Unity editor Mono EditMode (GoldenCheckTests, in a full run of 1,321 tests, 0 failed); WebGL IL2CPP in Chrome (Web-disabled, ?golden&smoke=5, PASS with the SMOKE OK line too); and a macOS IL2CPP Universal player on Apple silicon (arm64 slice, -golden, exit 0). The x86_64 slice, Windows and Linux players were not run then (the Player golden workflow checks them on a release).

The player needs an absolute -logFile path: with a relative one it wrote no log file in the check run. The editor commands are fine with relative paths.

A content change can move the content hash without moving a golden. Settlement names, for example, are content (the hash covers content/map/settlements_1948.yaml) but not state. On 2026-10-12 the owner ruled that the game settlements lod and ramla take their 1948 names, Lydda and Ramle, as Village Statistics 1945 lists them ("Lydda (Urban)", "Ramle (Urban)"). That commit changed only those two names and added a source to each. Both goldens and the five-turn smoke are unchanged: .NET 667 of 667, and the WebGL golden&smoke=5 check passes with the three values above. Only the content hash moved. A save made before it therefore loads with the "different game content" warning (GameSession.Load) and replays to the same state.

Any change to a golden hash is a deliberate rules or content change. Such a change could be to a rule, a system, content in content/, or the hash format. When it happens, update the manifest and every place above in the same commit, add a row to the history below, and say in the commit message why the hash moved. If a hash changes and you did not mean to change rules or content, treat it as a determinism bug.

Content windows

Records with from/to (chapters spec 3.2) are resolved by date through Content.At(date). The active lists are built once at load, one per segment between cut dates, so a turn does a binary search and no allocation. When the date crosses a cut, ContentWindows.Update adds starting records at their defaults and removes ending ones, in ordinal id order, right after the date moves (before the cabinet, attention and events; the turn's effects already ran on the turn's start date). The hash reads the state keys, so it never sees a record outside its window. An event that fires late (after its window, EventSystem.Trigger) can find a record it names already gone: a condition on it is not met, and an option's effect or set_policy on it is left out. With no windows there is one segment and the same view everywhere.

Rebaseline history

Plan task Run New hash (first 8) Why
M3 Task 2 1950 d46ce5ff the Immigration budget line moves to the Prime Minister (the Ministry of Immigration ends on 1951-10-08), which renames a key in the hashed budget
M3 Task 3 1950 3d360deb CBS immigration for 1950 adds January-February 1950 arrivals inside the run
M3 Task 4 1950 e2fa3896 frumkin_commission (window from 1950-01-01) fires inside the run
M3 Task 5 1950 bf6baf3f economy calibrated (rules.yaml) and HistoricalAI sends reserves home after the armistices; capital_inflow_per_year 193.8 acts from 1948, so it moves this hash too
M3 Task 5 1952 1753fb88 new run covering the M3 content range. Task 4 fix round 1 changed 1951 event windows/effects (hula, forex, bonds, clothing rationing, education, reparations), so capital_inflow_per_year was re-fitted from 192 to 193.8 (first fx crisis in H2 1951)
M3 Task 5 fix 1 1950 43cc7773 capital_inflow_per_year re-fitted 193.8 -> 200 (it acts from 1948) after the missing $45m credit maturity was added
M3 Task 5 fix 1 1952 0359fa91 forex_crisis_1951 now applies the documented -45 fx (FRUS d513) in both options, and capital_inflow_per_year is 200
M3 Task 6 1950 549b49cb hash format adds the player's appointments (empty in the run); no state change
M3 Task 6 1952 860928f2 same
M3 Task 8 1950 f52fa804 hash format adds map placements (empty in the run); no state change
M3 Task 8 1952 d07cd3d2 same
M3 Task 8 fix 1950 ee5d0826 Task 8 fix: per-turn map-action counter in state (hash text gains mapturn=); no state change in the run
M3 Task 8 fix 1952 91bb6fb5 same
playtest 1 E 1950 a540b52e 1948 war-period events added (playtest 1 E): thirteen events whose windows open between 26 May and 31 December 1948; the second truce puts the run on monthly turns from 17 July to 17 October 1948, so the run ends 1950-02-27 after 35 turns (was 1950-02-26, turn 45)
playtest 1 E 1952 16c16d7e same; ends 1952-02-27 after 43 turns (was 1952-02-26, turn 53)
playtest 1 E fix 1 1950 d463d5b2 events gain requires: {event, option}; the hash text gains chosen{event=option;...} (the options each resolved event was resolved with); no change in the run
playtest 1 E fix 1 1952 f78fc68d same
playtest 1 E fix 1 1950 71e9904d 1948 war-period events revised after review: windows open on the turn before each decision, Yoav becomes the Hebron-hills decision, follow-ups use requires, Bernadotte's shared standing hit, security deltas halved, the elections a one-option card in January, two events added (Jerusalem cease-fire, Rhodes talks); still ends 1950-02-27, turn 35
playtest 1 E fix 1 1952 73590e32 same; still ends 1952-02-27, turn 43
Issue #2 1952 62f8ec61 sourced Yosef food-supply/rationing responsibility from 8 October 1951 replaces a vacancy and changes late-1951 rationing competence; 1950 and five-turn smoke unchanged
GDD Phase 1 PR 3 mvp-slice f6931e24 the slice gains its city (placeholder values) and the GDD turn: Money from jobs, housing, Social cohesion and arrivals, with the reference player filling the jobs (was 0451920a); the legacy goldens do not move
GDD Phase 1 PR 4 mvp-slice 6343eee8 training runs in the slice (step 4 of the GDD turn) and the reference player trains one halutz group on the first turn: it becomes a unit at Tel Aviv's hex on the fourth, its IL 0.5m adds to the debt, and the civilian it took frees a place that one more refugee is housed in (was f6931e24); the legacy goldens and scenario-mvp-absorb-first do not move
GDD Phase 1 PR 5 mvp-slice ec6f8c80 owner decision "Allow borrowing" (15 October 2026): training and pioneer work borrow what the Treasury lacks. The reference player's halutz group, ready on 15 September 1948, starts Mefalsim that turn (its IL 2m adds to the debt) and builds it on 15 March 1949; MvpGolden.Until moves from 15 January to 15 March 1949 (ten monthly turns) so the build finishes inside the run (was 6343eee8); the legacy goldens and scenario-mvp-absorb-first do not move
GDD Phase 2 PR 1 1950 f97e9f81 GDD v0.1: monthly turns and the GDD loop on (spec 2026-10-12 section 3 step 5). The real content runs systems.loop: gdd with monthly turns on every rung, and the golden runs play the reference player, so the run ends 1950-01-15 after 20 turns (was 1950-02-27, turn 35) and gains the GDD sections (city{}, coh=, units[], next=: the halutz group trained from 15 August 1948; site{mefalsim=done,1949-06-15}: the build of #566, with its IL 2m in the debt). Event windows re-timed per #567 (idf_founding, altalena, jerusalem_military_government, latrun_proposal, operation_horev, elections_1949), and PR 1a (#577) merged in: the five war cards rewritten (options keep and stop_at_ceasefire) and the sensitive options' gains replaced by costs (morale and International Support lower); the Phase 1 and #566 values in content as design values, unchanged; the FRUS anchor holds with no refit; mvp-slice (ec6f8c80) retired; #315 merged earlier (#475) and moved no golden (was 71e9904d)
GDD Phase 2 PR 1 1952 75d38299 same; the game now ends on 15 January 1952 (R-LOOP-004), the first turn on or after 1 January 1952: ends 1952-01-15, turn 44, over=mvp_ended (was 1952-02-27, turn 43, still playing); the reparations claim's cost moves to Money (treasury -1, #577), so the debt is IL 1m higher (was 62f8ec61)
GDD Phase 2 PR 1 mvp-end 75d38299 new: the reference player over the whole game, to 15 January 1952; the same run as hash1952
GDD Phase 2 PR 1 smoke5 85191d48 five monthly turns end on 1948-10-15 (was 1948-06-19); the smoke answers cards while the turn's Attention pays for them, so the third card of 15 July 1948 goes to its minister (was d3fdd459)
GDD Phase 2 PR 1 scenario-cautious 58eb01db monthly turns: ends 1949-01-15, turn 8 (was 1949-01-02, turn 23); rhodes_talks and elections_1949 now fire inside the run; the GDD sections appear; #577's rewritten cards and costs (was 1d5b8d71)
GDD Phase 2 PR 1 scenario-reckless e2f409a5 same (was c0bd12bc)
GDD Phase 2 PR 1 scenario-mvp-absorb-first f8858f6e plays the real content (the overlay is retired, content: mvp dropped); the Horev window edit makes rhodes_talks and sinai_withdrawal fire; #577's rewritten cards and costs (was 59ea3181)
GDD Phase 1 PR 6 1950 ad806776 the first new yishuv objective (R-OBJ-001 to R-OBJ-003), rebased onto the cutover: the reference player builds Mefalsim on 15 June 1949, the month of its 12 June 1949 founding, so the objective completes on time for 100 points (obj{first_new_yishuv=1949-06-15}, score=100) and its card fires and is answered (first_new_yishuv=continue); no other event moves, so the card displaces nothing under the soft-event cap; objective_points, objective_bonus_per_month and objective_bonus_cap enter rules.yaml as design values (D-B6, D-B2) (was f97e9f81)
GDD Phase 1 PR 6 1952 0f54dc51 same (was 75d38299)
GDD Phase 1 PR 6 mvp-end 0f54dc51 same; the same run as hash1952 (was 75d38299); smoke5 and the three scenarios do not move
GDD Phase 2 PR 3 1950 ca953c98 the national Food stock (R-FOOD-001 to R-FOOD-003; maintainer decision of 16 October 2026, #382, option (b)): a new hashed line food=stock,need,unmet, written only when set. The stock starts at cover_months (3) of the need and the monthly import keeps it there, so its food_supply term is exactly 0 and nothing else in the run moves (food=3003,1001,0) (was ad806776)
GDD Phase 2 PR 3 1952 b42380ab same, and the 1951 currency crisis now cuts Food imports into the stock (Q-POOL-4) instead of shifting food_supply directly: the legacy cut is gone (shift food_supply 0.15, was 0.1167: the events' 0.15 less the cut's 0.0333 over the two crisis turns), the last turn imports nothing, the stock falls to 2732 against a full 4170 (food=2732,1390,0), and the stock term takes 0.0345 off, so food_supply ends at 0.8555 (was 0.8567); the voter groups read food_supply, so stab moves with it (0.5098, was 0.5095) (was 0f54dc51)
GDD Phase 2 PR 3 mvp-end b42380ab same; the same run as hash1952 (was 0f54dc51)
GDD Phase 2 PR 3 smoke5 848969ce the new food= line only; the five turns hold a full stock (was 85191d48)
GDD Phase 2 PR 3 scenario-cautious 5553c31a the new food= line only (food=2253,751,0) (was 58eb01db)
GDD Phase 2 PR 3 scenario-reckless 42041362 the new food= line only (food=2253,751,0) (was e2f409a5)
GDD Phase 2 PR 3 scenario-mvp-absorb-first e01bb37a the new food= line only (food=2556,852,0) (was f8858f6e)
GDD Phase 3 PR 1 1950 ea730229 GDD Phase 3 PR 1's twelve cities (#821, R-CITY-006, R-CITY-007): the city{} section holds the twelve census starts (511 units; Tel Aviv 248 civilians and 106 employed, was 100 and 80) with every job filled, and spending_per_capita_per_year is re-derived with their job income (0.0001036, was 0.000093; owner, 16 October 2026, option (a)). The reference player sends no AssignWorkers on the first turn, so it trains its halutz group on 15 May 1948 (was 15 August) and builds Mefalsim on 15 March 1949 (was 15 June): score=115 with the early bonus (was 100) (was ca953c98)
GDD Phase 3 PR 1 1952 ca887e45 same; debt ends at 170.9 (was 157.0) and coh at 0.4213 (was 0.4521) (was b42380ab)
GDD Phase 3 PR 1 mvp-end ca887e45 same; the same run as hash1952 (was b42380ab)
GDD Phase 3 PR 1 smoke5 4104103f the twelve cities and the new spending in the first five turns (was 848969ce)
GDD Phase 3 PR 1 scenario-cautious 5f162f0f the twelve cities and the new spending (was 5553c31a)
GDD Phase 3 PR 1 scenario-reckless a10877e2 same (was 42041362)
GDD Phase 3 PR 1 scenario-mvp-absorb-first 53852f2c same (was e01bb37a)
GDD Phase 3 PR 2 1950 d22b369f GDD Phase 3 PR 2's arrivals over the cities (R-CITY-003, Q-ARRV-1, Q-ARRV-3): the cities take arrivals from the census day, 8 November 1948, not 15 May, so 78 fewer units arrive in them (269 refugees waiting on 1 January 1950, was 347), and they are spread over the twelve cities by census people instead of all landing in Tel Aviv (Tel Aviv 131, Haifa 52, Jerusalem 44). Only the city{} section and coh move (0.5418, was 0.5106); the national head count, the debt and the reference player's orders do not (was ea730229)
GDD Phase 3 PR 2 1952 21029217 same; 608 refugees waiting at the end (was 686), Tel Aviv 296 of them; coh 0.4364 (was 0.4213) (was ca887e45)
GDD Phase 3 PR 2 mvp-end 21029217 same; the same run as hash1952 (was ca887e45)
GDD Phase 3 PR 2 smoke5 3f7a0fe0 the first five turns end before the census day, so the cities take no arrivals in them (was 4104103f)
GDD Phase 3 PR 2 scenario-cautious a9afb0ad the arrivals over the cities; city{} and coh only (was 5f162f0f)
GDD Phase 3 PR 2 scenario-reckless 8f209d56 same (was a10877e2)
GDD Phase 3 PR 2 scenario-mvp-absorb-first 8601da93 same (was 53852f2c)
GDD Phase 2 PR 4 1950 2766b45b Social cohesion is morale (D-C1, Reading 2; R-CITY-005, R-NAT-001): the coh= line goes (a hash format change), and two new effect-web sources of morale, work_lifts_cohesion (the cities' employed share) and food_lifts_cohesion (food_supply), k 0.10 and inertia 2 each, act every turn (Q-NAT-2). morale ends at 0.5545 (was 0.4255); approval reads it (Q-NAT-5), so stab ends at 0.5612 (was 0.5123). Two new situations, low_cohesion and low_intl_support, do not fire in the run (was d22b369f)
GDD Phase 2 PR 4 1952 f8cd22d7 same; morale ends at 0.5691 (was 0.4388), stab at 0.5683 (was 0.5098) (was 21029217)
GDD Phase 2 PR 4 mvp-end f8cd22d7 same; the same run as hash1952 (was 21029217)
GDD Phase 2 PR 4 smoke5 2436927c the coh= line goes and the two new sources of morale (was 3f7a0fe0)
GDD Phase 2 PR 4 scenario-cautious 28fd1405 same; morale ends at 0.6183 (was 0.5296) (was a9afb0ad)
GDD Phase 2 PR 4 scenario-reckless af7b2abd same (was 8f209d56)
GDD Phase 2 PR 4 scenario-mvp-absorb-first b16a8740 same (was 8601da93)
GDD Phase 4 PR 1 1950 37c51050 the MVP's four yishuv sites (R-SITE-004, Q-SITE-1, #785): the reference player builds them in ordinal id order, so Beit Kama is built on 1949-03-15 and Gilat on 1949-09-15, and Mefalsim is still being built on 1 January 1950. The first-new-yishuv objective and its card leave the run (score 0, was 115), and the debt is IL 4m higher for the extra builds (was 2766b45b)
GDD Phase 4 PR 1 1952 7745031b same; Mefalsim is built on 1950-03-15 and Re'im on 1950-09-15, so the objective completes nine months after its reference date with no timing bonus (score 100, was 115); the debt is IL 6m higher (was f8cd22d7)
GDD Phase 4 PR 1 mvp-end 7745031b same; the same run as hash1952 (was f8cd22d7)
GDD Phase 4 PR 6 1950 ef765972 the veteran pioneer (R-VET-001, R-VET-002, D-D9, #790): the reference player's halutz group works the six turns of the Beit Kama build (still built on 1949-03-15), becomes a veteran_halutz with 6 turns of service (1:veteran_halutz@-3,19/6) and builds the other three sites in 3 turns each: Gilat on 1949-06-15 (was 1949-09-15), Mefalsim on 1949-09-15 and Re'im on 1949-12-15. The first-new-yishuv objective completes on 1949-09-15 and its card fires and is answered in the run (score 100, was 0), and the debt is IL 2m higher for the extra builds (was 37c51050)
GDD Phase 4 PR 6 1952 7b4bffe0 same; Mefalsim is built on 1949-09-15 (was 1950-03-15) and Re'im on 1949-12-15 (was 1950-09-15). The objective still completes after its reference date with no timing bonus, so the score (100) and the debt do not change (was 7745031b)
GDD Phase 4 PR 6 mvp-end 7b4bffe0 same; the same run as hash1952 (was 7745031b)
GDD Phase 4 PR 3 1950 38c8a484 build a training camp (R-CAMP-001, #787): once the four yishuv sites are taken, the reference player's veteran group builds the camp sites in ordinal id order. At the end of the 1950 run (1950-01-15) Camp 80 is being built (camp_80=b,1,4), and the debt is IL 2m higher (was ef765972)
GDD Phase 4 PR 3 1952 1e21fcd3 same; at 2 turns of progress a turn (R-VET-002) Camp 80 is built on 1950-03-15 and Tzrifin on 1950-06-15. The debt is IL 4m higher (2 x IL 2m). The reference player trains no group after the first, so the new facilities (a military camp each for Hadera and Rishon LeZion, owner "Both military" on #927) change nothing else (was 7b4bffe0)
GDD Phase 4 PR 3 mvp-end 1e21fcd3 same; the same run as hash1952 (was 7b4bffe0)
GDD Phase 4 PR 4 1950 e4ebff9b plow fields (R-FIELD-001, R-FIELD-002, #788), rebased on the training camps (#927): once the four yishuv sites are taken, the reference player's veteran group plows their fields in ordinal id order, before any camp. At the end of the 1950 run (1950-01-15) Beit Kama's field is still being plowed (beit_kama_field=b,1,1) instead of Camp 80 being built, and the debt is IL 1.5m lower (IL 0.5m for the field instead of IL 2m for the camp) (was 38c8a484)
GDD Phase 4 PR 4 1952 e0d8212a same; at 2 turns of progress a turn (R-VET-002) the fields are plowed on 1950-02-15 (Beit Kama), 1950-04-15 (Gilat), 1950-06-15 (Mefalsim) and 1950-08-15 (Re'im), so Camp 80 is built on 1950-11-15 (was 1950-03-15) and Tzrifin on 1951-02-15 (was 1950-06-15). The debt is IL 2m higher (4 x IL 0.5m), and the Food stock ends 4 units higher (2736, was 2732), which lifts food_supply and stability in the fourth decimal (was 1e21fcd3)
GDD Phase 4 PR 4 mvp-end e0d8212a same; the same run as hash1952 (was 1e21fcd3)
GDD Phase 3 PR 3 1950 7cf678b6 GDD Phase 3 PR 3's housing (R-CITY-004, R-HOUSE-001 to R-HOUSE-004; Q-HOUSE-1, Q-HOUSE-2, Q-HOUSE-5, Q-START-3; owner, 17 October 2026, Q-PR3-A (a)), re-pinned on main after #927 and #871: no hard cap, so every refugee is housed on the next housing step (18 refugees waiting on 1 January 1950, the last turn's arrivals, was 269), the people beyond a city's housing are its overflow, and each city's housing condition (start 0.35) falls by 0.15 times its overflow share a turn; 11 of the 12 cities are at condition 0 by then (Rosh Pinna, with no overflow, keeps 0.35). The condition joins each city{} record. The national housing indicator's base is now the cities' people-weighted mean condition, so housing falls to 0.0850 (was 0.4345) and housing_crisis is on; through homes_lift_morale, Social cohesion (morale) falls to 0.4990 (was 0.5545), with work_lifts_cohesion lower too because more civilians share the same jobs; stability 0.4970 (was 0.5612). The national head count, the Treasury, the debt, Food, the camp and field builds and the reference player's orders do not move (was e4ebff9b)
GDD Phase 3 PR 3 1952 94969472 same; 8 refugees waiting at the end (was 608), every city at condition 0; housing 0.0850 (was 0.4350), housing_crisis on, morale 0.4946 (was 0.5691), stability 0.4487 (was 0.5683); the housing shift reads -1.4e-17 (was 0), IEEE rounding of the level minus the moving inner value (was e0d8212a)
GDD Phase 3 PR 3 mvp-end 94969472 same; the same run as hash1952 (was e0d8212a)
GDD Phase 3 PR 3 smoke5 3005405d the condition joins each city{} record; the first five turns have no overflow, so housing and morale are unchanged (was 2436927c)
GDD Phase 3 PR 3 scenario-cautious 6507ddc3 the housing rules: city{}, housing, morale, its two links and stability (was 28fd1405)
GDD Phase 3 PR 3 scenario-reckless 8041988a same (was af7b2abd)
GDD Phase 3 PR 3 scenario-mvp-absorb-first a77a8f20 same (was b16a8740)
GDD Phase 4 PR 5 1952 f9f47f5c dry swamp and the Hula event (R-SWAMP-001, R-SWAMP-002, #789), rebased on the housing condition (#855) and the alternate-timeline stack (#960, #966, #972): hula_drainage now fires only while the player's works at the Hula swamp site are in progress, and the reference player drains no swamp, so the card leaves fired[] and chosen{} (it fired on 1951-03-15 with continue_works). Without that choice's effects security ends at 0.5061 (was 0.5361) and intl_standing at 0.83 (was 0.78), which moves morale (0.4916, was 0.4946) and stability (0.4432, was 0.4487) through the effects web (was 94969472)
GDD Phase 4 PR 5 mvp-end f9f47f5c same; the same run as hash1952 (was 94969472)
GDD Phase 4 PR 1 (Kiryat Shmona, Beit Shemesh) 1950 69ac1384 Kiryat Shmona and Beit Shemesh join the yishuv sites (Q-SITE-1, owner, #782 and #785), re-pinned on main after #871, #855 and #968. The reference player takes the yishuv sites in ordinal id order before any field or camp, so its veteran builds Beit Shemesh on 1949-06-15, Gilat on 1949-09-15 and Kiryat Shmona on 1949-12-15, and is building Mefalsim at the cut (mefalsim=b,1,4; it was plowing Beit Kama's field). The first-new-yishuv objective no longer completes inside the run, so its card does not fire and the score is 0 (was 100); the debt is IL 1.5m higher (IL 2m for the build in progress instead of IL 0.5m for the field) (was 7cf678b6)
GDD Phase 4 PR 1 (Kiryat Shmona, Beit Shemesh) 1952 b34faec6 same; Mefalsim is built on 1950-03-15 (was 1949-09-15) and Re'im on 1950-06-15 (was 1949-12-15). The six fields are plowed after them, in ordinal id order: Beit Kama 1950-08-15, Beit Shemesh 1950-10-15, Gilat 1950-12-15, Kiryat Shmona 1951-02-15, Mefalsim 1951-04-15 and Re'im 1951-06-15, so Camp 80 is built on 1951-09-15 (was 1950-11-15) and Tzrifin on 1951-12-15 (was 1951-02-15). The objective completes on 1950-03-15, after its reference date, so the score (100) does not change. The debt is IL 5m higher (2 x IL 2m for the yishuvim, 2 x IL 0.5m for their fields), and the Food stock ends 2 units higher (2738, was 2736), which lifts food_supply in the fifth decimal and stability in the sixth (was f9f47f5c)
GDD Phase 4 PR 1 (Kiryat Shmona, Beit Shemesh) mvp-end b34faec6 same; the same run as hash1952 (was f9f47f5c)
GDD Phase 3 PR 4 1950 9922495a city projects (R-PROJ-001 to R-PROJ-003, #735; Q-PROJ-1 to Q-PROJ-5, Q-HOUSE-3, owner, 15 October 2026, #375): the reference player's stated policy (D-A7) now starts a housing project, one at a time, in the city with the most overflow, on the first free hex next to it, borrowing as its other orders do. Its first start is on 1949-03-15 in Tel Aviv, built on 1949-08-15 (+5 capacity, +0.03 condition); the second is being built. cproj[...] joins the state; mapturn now counts the turn's project starts. Tel Aviv's capacity 253 (was 248) and condition 0.0168 (was 0), so housing 0.0932 (was 0.0850), morale 0.5018 (was 0.4990), stability 0.5002 (was 0.4970); debt IL 91.5m (was 79.5m, 2 x IL 6m); rebased on Kiryat Shmona and Beit Shemesh (#944) and the event city effect (#1001) (was 69ac1384)
GDD Phase 3 PR 4 1952 548a6acc same; five Tel Aviv projects built (1949-08-15 to 1951-08-15; capacity 273, was 248) and a sixth, in Haifa, being built. Every city is back at condition 0 by the end, so housing is unchanged and morale moves only in the 10th decimal; debt IL 223.9m (was 187.9m, 6 x IL 6m), inside invariants.yaml's 100 to 300 band; stability 0.4432 (unchanged to 4 places) (was b34faec6)
GDD Phase 3 PR 4 mvp-end 548a6acc same; the same run as hash1952 (was b34faec6)
GDD Phase 7 PR 5 scenario-mvp-defeat d1e0dcf6 new: full mobilisation and the non-historical option of the four foreign currency cards; the debt passes the IL 250m ceiling on 15 April 1951, foreign currency runs out on 15 August, food imports stop, the stock is empty from 15 November, and the third turn end with Money and Food both exhausted ends the game with defeat_economy on 15 January 1952 (exhausted=3). No other golden moves: the reference and the no-order runs never have Food at 0, so the counter stays at 0 and out of the hash. Rebased on the Hula swamp (#968): hula_drainage no longer fires in this run (no works at the Hula) and, as in the 1952 run on main, iraqi_exodus resolves to airlift (was pace), so the end state differs; the defeat dates above are unchanged (was cbebedfa before the rebase)
GDD Phase 3 PR 5 scenario-mvp-absorb-first be13c409 the job default and the craftsman (R-CITY-009, R-CRAFT-001, R-CRAFT-002; Q-JOBS-1, Q-CRAFT-1 to Q-CRAFT-3, owner, 15 October 2026, #375; #736): the scenario's one order, AssignWorkers for 90 of Tel Aviv's 106 jobs on 15 May 1948, now turns Tel Aviv's job default off, so its city{} record gains ,jobs=manual. No number moves: Tel Aviv keeps its 90 as before, and the other cities' jobs stay full. hash1950, hash1952, mvp-end, smoke5, scenario-cautious, scenario-reckless and scenario-mvp-defeat do not move: the reference player no longer sends AssignWorkers, and on those runs the default fills the same jobs it did; nobody trains a craftsman (was a77a8f20)
#457 item 9 (official records) 1952 722d90a2 owner ruling "Follow the official records" (18 October 2026, #457): education_registration_1951 opens on 12 February 1951, its first cited plenum record (was 29 November 1950), so the card fires on 15 February 1951 instead of 15 December 1950 and its morale cost lands two turns later. At the end only coalition stability moves (0.443222 to 0.443592). The same PR's Education vacancy from 19 May to 8 October 1951 (item 8) and the 14 May 1948 start (item 2, unchanged) move no golden: measured with the event date as before (was 548a6acc)
#457 item 9 (official records) mvp-end 722d90a2 same; the same run as hash1952 (was 548a6acc)
#457 item 9 (official records) scenario-mvp-defeat 19143463 same card date; at the end only coalition stability moves (0.457037 to 0.457407), and the scenario still ends in defeat_economy on 15 January 1952 (was d1e0dcf6)
GDD Phase 2 PR 3b 1950 4f1202cf the national Arms pool (R-ARMS-001 to R-ARMS-003, #781; owner, 18 October 2026: "Approve as drafted"): the new arms= line, 12 at the start. The reference player trains no recruits and buys no Arms, so it stays 12 and nothing else moves (was 9922495a)
GDD Phase 2 PR 3b 1952 28334950 same; arms=12 at the end (was 722d90a2)
GDD Phase 2 PR 3b mvp-end 28334950 same; the same run as hash1952 (was 722d90a2)
GDD Phase 2 PR 3b smoke5 8d000459 same; arms=12 (was 3005405d)
GDD Phase 2 PR 3b scenario-cautious a6108c91 same; no scenario trains recruits or buys Arms, so arms=12 (was 6507ddc3)
GDD Phase 2 PR 3b scenario-reckless a0f094a1 same (was 8041988a)
GDD Phase 2 PR 3b scenario-mvp-absorb-first dee4f039 same (was be13c409)
GDD Phase 2 PR 3b scenario-mvp-defeat e71076f7 same; the defeat run is unchanged apart from arms=12 (was 19143463)
GDD Phase 7 PR 4 1950 25ea8fbf the ministers on the GDD systems (R-MIN-001 to R-MIN-003; Q-MIN-1 to Q-MIN-3, owner, 17 and 18 October 2026, #784; #808): construction, pioneer and field work and absorption run at their office's multiplier, with the build's fractional carry kept on the record (an in-progress site writes it, gilat=b,1,4,0.652...). At the cut the veteran group has built Beit Kama on 1949-06-15 (was 1949-03-15) and Beit Shemesh on 1949-11-15 (was 1949-06-15) and is on Gilat, so two fewer yishuvim are borrowed and the debt is IL 87.5m (was 91.5m); slower absorption houses fewer civilians, so the cities' overflow is smaller and housing 0.1062 (was 0.0932) and morale 0.5050 (was 0.5018) rise (was 4f1202cf)
GDD Phase 7 PR 4 1952 ef8c5fea same; the veteran's builds take 5 turns at 2 x 0.624 = 1.248 a turn, so Gilat is built on 1950-04-15, Kiryat Shmona on 1950-09-15, Mefalsim on 1951-02-15 and Re'im on 1951-07-15 (were 1949-09-15, 1949-12-15, 1950-03-15 and 1950-06-15), only two fields are plowed (Beit Kama 1951-10-15, Beit Shemesh 1952-01-15) and the camps never start. The debt is IL 211.9m (was 223.9m) and the Food stock 2734 (was 2738). The first-new-yishuv objective completes on 1951-02-15, after its reference date, so the score stays 100 (was 28334950)
GDD Phase 7 PR 4 mvp-end ef8c5fea same; the same run as hash1952 (was 28334950)
GDD Phase 7 PR 4 scenario-cautious 9a3d241e the absorption and construction multipliers: the cities' housed counts and conditions move, and housing, morale and stability with them; no order or date changes (was a6108c91)
GDD Phase 7 PR 4 scenario-reckless d7dbcf75 same; each city houses fewer of its refugees a turn, so more wait (was a0f094a1)
GDD Phase 7 PR 4 scenario-mvp-absorb-first 234e32ee same; the scenario's one order stands, and every city's housed count moves (was dee4f039)
GDD Phase 7 PR 4 scenario-mvp-defeat f3daaa70 same; only the housing web's numbers move, in the third decimal and below - the defeat is unchanged, still defeat_economy on 15 January 1952 (was e71076f7)
Phase 7 PR 4 carry canonical fix 1950 e3c76f5f Nonzero site and city-project carry is written as IEEE bits, consistently on .NET and Unity Mono, instead of runtime-dependent round-trip decimal text. Only canonical representation changes; the stored progress, all other snapshot lines, completion dates and gameplay are unchanged (was 25ea8fbf).
Phase 7 PR 4 carry canonical fix 1952 37866907 Nonzero site and city-project carry is written as IEEE bits, consistently on .NET and Unity Mono, instead of runtime-dependent round-trip decimal text. Only canonical representation changes; the stored progress, all other snapshot lines, completion dates and gameplay are unchanged (was ef8c5fea).
Phase 7 PR 4 carry canonical fix mvp-end 37866907 Nonzero site and city-project carry is written as IEEE bits, consistently on .NET and Unity Mono, instead of runtime-dependent round-trip decimal text. Only canonical representation changes; the stored progress, all other snapshot lines, completion dates and gameplay are unchanged (was ef8c5fea).
#738 remaining scenario step scenario-mvp-absorb-first c045ea90 extends the run to 15 January 1950 so it completes absorb_immigrants (R-OBJ-004): two halutz cohorts free city housing, fourteen housing projects take the overflow and two industry projects raise employment, using only existing commands (was 234e32ee)
GDD Phase 4 PR 7 (#791) 1952 c17ac550 The first-new-yishuv objective becomes the all-six yishuvim infrastructure part, recorded 15 July 1951 instead of 15 February 1951. Neither corridor completes in the reference run, so the shared 100-point base is not paid: score 0 instead of 100. Only the objective record/date and score change; hash1950, smoke5 and all scenario goldens stay unchanged (was 37866907).
GDD Phase 4 PR 7 (#791) mvp-end c17ac550 Same full run and changes as hash1952 (was 37866907).
GDD Phase 5 PR 4: the sourced starting forces enter the state 1950 259133fc A new game starts with the 12 forces of map/units_1948.yaml (R-UNIT-006): seven IDF brigades as veteran_soldier and the five Arab forces with a row, ids 0 to 11, so the units[ section gains them and every later unit, cohort and city project id moves up by 12 (the halutz group 1 is now 13, next= 4 is 16). Nothing else in the snapshot moves: no rule reads the forces yet (was e3c76f5f)
GDD Phase 5 PR 4: the sourced starting forces enter the state 1952 452c65ec same; next= 7 is 19 (was 37866907)
GDD Phase 5 PR 4: the sourced starting forces enter the state mvp-end 452c65ec same as 1952 (was 37866907)
GDD Phase 5 PR 4: the sourced starting forces enter the state smoke5 e5a66f9c same; the 12 forces are in the state from the first turn (was 8d000459)
GDD Phase 5 PR 4: the sourced starting forces enter the state scenario-cautious b4e5caa1 same; the scenario has no unit of its own, so it gains the units[ section and next=12 (was 9a3d241e)
GDD Phase 5 PR 4: the sourced starting forces enter the state scenario-reckless 8d644f6e same as cautious (was d7dbcf75)
GDD Phase 5 PR 4: the sourced starting forces enter the state scenario-mvp-absorb-first 7d37c0d0 same as cautious; the longer #738 run now also carries the 12 forces (was c045ea90)
GDD Phase 5 PR 4: the sourced starting forces enter the state scenario-mvp-defeat 7bd5158f same as cautious; the defeat is unchanged (was f3daaa70)
GDD Phase 5 PR 8: combat (#747) 1950 73eb50db The reference player's attack policy (attack a neighbouring enemy hex at 3:2 or better) sends unit 0 (strength 5, 10,9) against the Syrian unit 9 (3, the valley's 10,10) from 15 May 1948. Four engagements, one RNG draw each, 15 May to 15 August 1948: unit 0 ends at strength 3, unit 9 is destroyed and leaves units[, and the Arms stock falls from 12 to 8 (1 an attack). The RNG state moves from the seed for the first time in the run, since nothing else in it draws. Nothing else in the snapshot moves; the scenarios and smoke5 give no attack and are unchanged (was 259133fc)
GDD Phase 5 PR 8: combat (#747) 1952 d9922fc9 same (was 452c65ec)
GDD Phase 5 PR 8: combat (#747) mvp-end d9922fc9 same as 1952 (was 452c65ec)
Combined #1186 + #791 (C20) 1952 72b8e8014c96c943e390f0fd109df68b78f59e5560b05c2d231e1fbe599d1280 Keeps all 12 starting forces and their shifted ids from #1186, plus the three-part infrastructure objective from #791. Relative to main, only the objective record moves from first_new_yishuv=1951-02-15 to yishuvim=1951-07-15 and the separate 100-point base is no longer paid; neither corridor completes. All other snapshot lines, hash1950, smoke5 and scenario goldens are unchanged (was 452c65ece470e844f6898151c4d9328aa0b2e59df82db4763214d43a8e8ea7f1).
Combined #1186 + #791 (C20) mvp-end 72b8e8014c96c943e390f0fd109df68b78f59e5560b05c2d231e1fbe599d1280 The same full run and combined #1186 + #791 state as hash1952 (was 452c65ece470e844f6898151c4d9328aa0b2e59df82db4763214d43a8e8ea7f1).
GDD Phase 4 PR 7 (#791) on combat (#747) 1952 b389b278 #1226 merged with main after combat (#747). As in the rows above, the first-new-yishuv objective becomes the all-six yishuvim infrastructure part, recorded 15 July 1951 instead of 15 February 1951, and neither corridor completes, so the shared 100-point base is unpaid and the snapshot's score=100 line goes (a zero score is not written). Only those two obj{/score lines move; hash1950, smoke5 and the four scenario goldens equal main (was d9922fc9)
GDD Phase 4 PR 7 (#791) on combat (#747) mvp-end b389b278 same as 1952 (was d9922fc9)
#1240: the seven IDF rows re-sourced to Ben-Gurion's diary of 15 May 1948, owner 21 Oct 2026 1950 a1ad4211 The seven IDF starting rows take the diary's 15 May strengths, so the starting units move: Golani 5 -> 4, Alexandroni 5 -> 4, Kiryati 1 -> 3, Givati 5 -> 3, Negev 2 -> 1 (Carmeli 2 and Harel 1 unchanged). Unit 0 (Golani, 4 on 10,9) against the Syrian 3 on 10,10 is now 133 percent, under the reference player's 150, so the run has no engagement: unit 0 ends at 4, not 3, the Syrian unit 9 stays in units[, the Arms stay at 12 (was 8), and the RNG is never drawn, so the snapshot's header shows the seed, 1948. Nothing else in the snapshot moves (was 73eb50db)
#1240: the seven IDF rows re-sourced to Ben-Gurion's diary of 15 May 1948, owner 21 Oct 2026 1952 9cbd73ed same (was b389b278)
#1240: the seven IDF rows re-sourced to Ben-Gurion's diary of 15 May 1948, owner 21 Oct 2026 mvp-end 9cbd73ed same as 1952 (was b389b278)
#1240: the seven IDF rows re-sourced to Ben-Gurion's diary of 15 May 1948, owner 21 Oct 2026 smoke5 ed761c3b same starting strengths; the smoke run gives no attack (was e5a66f9c)
#1240: the seven IDF rows re-sourced to Ben-Gurion's diary of 15 May 1948, owner 21 Oct 2026 scenario-reckless 574c8a19 same starting strengths; only those five units[ lines move (was 8d644f6e)
#1240: the seven IDF rows re-sourced to Ben-Gurion's diary of 15 May 1948, owner 21 Oct 2026 scenario-cautious 00ec7ec2 same as reckless (was b4e5caa1)
#1240: the seven IDF rows re-sourced to Ben-Gurion's diary of 15 May 1948, owner 21 Oct 2026 scenario-mvp-absorb-first ac5f51bb same as reckless (was 7d37c0d0)
#1240: the seven IDF rows re-sourced to Ben-Gurion's diary of 15 May 1948, owner 21 Oct 2026 scenario-mvp-defeat f2ad1114 same as reckless; the defeat is unchanged (was 7bd5158f)
GDD Phase 5 PR 10: the enemy AI, part 1 (#749) 1950 1392f3eb Resolver step 1 gives the enemy its orders (R-ENEMY-001 to R-ENEMY-004). On 15 May and 15 June 1948 the four scripted forces march on their 11 June anchors (Egypt 7 to -8,25, one hex short of Isdud, round Yad Mordechai and the Negev brigade; the Arab Legion 8 to Latrun, -2,22; Syria 9 to 13,6, one short of the Mishmar HaYarden bridgehead; Iraq 10 to Jenin, 6,13), and every enemy unit fortifies from then on (order=fortify). The Syrian unit leaves Golani's side on the first turn (13,6 by the second), so no Israeli attack is ordered there; since #1240 none was (Golani's 4 against the Syrian 3 is under the policy's 3:2), and the Arms stay at 12. No enemy unit attacks on a turn whose middle day is inside a truce (R-ENEMY-006: 15 June to 15 September 1948). On the turn of 15 October 1948 the Legion attacks Harel beside Latrun at 4:1: one engagement, one draw, unit 5 is destroyed and the Legion is at 4. It is the run's only engagement, so the RNG state moves off the seed the header showed (was a1ad4211)
GDD Phase 5 PR 10: the enemy AI, part 1 (#749) 1952 cc96cd0a same (was 9cbd73ed)
GDD Phase 5 PR 10: the enemy AI, part 1 (#749) mvp-end cc96cd0a same as 1952 (was 9cbd73ed)
GDD Phase 5 PR 10: the enemy AI, part 1 (#749) smoke5 11c41014 the enemy's step 1 orders from the first turn, as in 1950. The run ends on 15 October 1948, inside the truce guard's reach (R-ENEMY-006), so Harel survives it: the Legion fortifies on the turns of 15 June to 15 September 1948 (was ed761c3b)
GDD Phase 5 PR 10: the enemy AI, part 1 (#749) scenario-cautious 230791a1 the units only, as in 1950: the scripted marches, the fortify orders and the Legion's engagement with Harel (was 00ec7ec2)
GDD Phase 5 PR 10: the enemy AI, part 1 (#749) scenario-reckless d6fd9e89 same as cautious (was 574c8a19)
GDD Phase 5 PR 10: the enemy AI, part 1 (#749) scenario-mvp-absorb-first de176931 same as cautious (was ac5f51bb)
GDD Phase 5 PR 10: the enemy AI, part 1 (#749) scenario-mvp-defeat 7404295a the units as in cautious; the engagement's draw comes before the delegated events' draws, so a delegated iraqi_exodus resolves to pace (was airlift), and housing, international standing, morale and stability follow. The defeat is unchanged (was f2ad1114)
#1241: Etzioni starts, the 7th Brigade arrives (R-UNIT-009), on the enemy AI (#1263) 1950 d14a4899 Etzioni's sourced 15 May strength (3,166, Ben-Gurion's diary) starts it under R-UNIT-006 as unit 5 (strength 3, Jerusalem's 0,23), so Harel and every later starting force take the next id (Harel 6, Egypt 8, the Legion 9, Syria 10). The 7th Brigade's reserved id 13 is the first after the starting rows; it arrives on the 15 June 1948 turn (its row's 24 May 1948 strength, 1,650 = 2, at Hulda's -4,23) and stands there. Every later id moves up by two (next=18). Nothing else in the snapshot moves: the enemy's scripted marches, the fortify orders and the Legion's 15 October 1948 engagement that destroys Harel are as on main, and the RNG and the Arms are unchanged (was 1392f3eb)
#1241: Etzioni starts, the 7th Brigade arrives (R-UNIT-009), on the enemy AI (#1263) 1952 3bacd2b1 same as 1950: the ids and the two units only (was cc96cd0a)
#1241: Etzioni starts, the 7th Brigade arrives (R-UNIT-009), on the enemy AI (#1263) mvp-end 3bacd2b1 same as 1952 (was cc96cd0a)
#1241: Etzioni starts, the 7th Brigade arrives (R-UNIT-009), on the enemy AI (#1263) smoke5 7559f30c same: the ids and the two units only (was 11c41014)
#1241: Etzioni starts, the 7th Brigade arrives (R-UNIT-009), on the enemy AI (#1263) scenario-reckless 030817e3 same: the ids and the two units only (was d6fd9e89)
#1241: Etzioni starts, the 7th Brigade arrives (R-UNIT-009), on the enemy AI (#1263) scenario-cautious e45c5429 same as reckless (was 230791a1)
#1241: Etzioni starts, the 7th Brigade arrives (R-UNIT-009), on the enemy AI (#1263) scenario-mvp-absorb-first 64bdd93b same as reckless (was de176931)
#1241: Etzioni starts, the 7th Brigade arrives (R-UNIT-009), on the enemy AI (#1263) scenario-mvp-defeat 54ec7d6b same as reckless; the defeat is unchanged (was 7404295a)
GDD Phase 5 PR 15 slice 4 (#754) smoke-orders 98422be8 new: one turn from the golden seed whose only order is unit 0 (Golani) moving from 10,9 to 7,11, given through the unit card in the browser run; no other golden moves
GDD Phase 5 PR 11, slice 1: the call-up retires in the GDD loop (#750, #1148) 1950 f70d7cc2 R-WAR-002: the GDD loop refuses the call-up order, so HistoricalAI sends none, and the call-up follows the rung down to its minimum the turn an event lowers it: 40% on the first truce's rung 4 (15 August 1948) and 5% on the armistices' rung 2 (15 April 1949), each a turn before HistoricalAI's order used to. Only the debt line moves, IL 87.5m to 81.8m. R-WAR-003: the ceasefire clock never starts, which this run's end state does not show (rung 2, cf=0 either way) (was d14a4899)
GDD Phase 5 PR 11, slice 1: the call-up retires in the GDD loop (#750, #1148) 1952 ca78893d same as 1950: only the debt line, IL 211.9m to 206.1m; the 1950-51 rise stays inside the Michaely anchor's tolerance (EconomyCalibrationTests) (was 3bacd2b1)
GDD Phase 5 PR 11, slice 1: the call-up retires in the GDD loop (#750, #1148) mvp-end ca78893d same as 1952 (was 3bacd2b1)
GDD Phase 5 PR 11, slice 1: the call-up retires in the GDD loop (#750, #1148) smoke5 71a4cef0 the smoke run sends no call-up order, so on main the 80% call-up stayed through the first truce; now it follows rung 4 down to 40%. Only mob (0.8 to 0.4) and the debt (IL 8.9m to 2.9m) move (was 7559f30c)
GDD Phase 5 PR 11, slice 1: the call-up retires in the GDD loop (#750, #1148) scenario-cautious d6e7cea0 the call-up follows the truce's rung down and back up (debt IL 27.0m to 21.0m), and the ceasefire clock never starts (cf=40 to cf=0) (was e45c5429)
GDD Phase 5 PR 11, slice 1: the call-up retires in the GDD loop (#750, #1148) scenario-reckless 57b3c29d the scenario's set_mobilization: 1.0 order is gone (the GDD loop refuses it), so the call-up is the war rung's 80% (mob 1 to 0.8, debt IL 35.0m to 21.0m), and cf=40 to cf=0 as in cautious (was 030817e3)
GDD Phase 5 PR 11, slice 1: the call-up retires in the GDD loop (#750, #1148) scenario-mvp-absorb-first 912504be the scenario sends no call-up order, so on main the 80% call-up stayed after the armistices; now it follows rung 2 down to 5% (mob 0.8 to 0.05, debt IL 316.4m to 272.9m) (was 64bdd93b)
GDD Phase 5 PR 11, slice 1: the call-up retires in the GDD loop (#750, #1148) scenario-mvp-defeat 109b0a5e the retired full call-up is replaced by 120 Arms bought on credit on the first turn (arms 12 to 132), and the call-up follows the armistices to 5% (mob 1 to 0.05). The debt passes the ceiling in July 1951 (was April) and ends at IL 281.6m (was 339.1m); the defeat is unchanged: defeat_economy on 15 January 1952 after three exhausted turns (was 54ec7d6b)
GDD Phase 5 PR 11, slice 1: the call-up retires in the GDD loop (#750, #1148) smoke-orders 4c991630 R-WAR-003: the ceasefire clock never starts, so the one-turn run ends on cf=0 (was cf=48); nothing else moves (was 98422be8)
GDD Phase 5 PR 3: recruit formations and their upkeep (#742, #1130) 1950 f3ec68b7 owner-accepted military upkeep (R-UNIT-010; IL 0.05m per Israeli formation per turn, owner 19 October 2026 on #1130) charges the starting forces and the 7th Brigade once it arrives; the reference player trains no recruits, so only debt moves, by IL 8.2m over 20 turns (164 formation-turns); every other snapshot field is unchanged (was f70d7cc2)
GDD Phase 5 PR 3: recruit formations and their upkeep (#742, #1130) 1952 6b374f69 same; debt alone rises by IL 17.8m over 44 turns (356 formation-turns) (was ca78893d)
GDD Phase 5 PR 3: recruit formations and their upkeep (#742, #1130) mvp-end 6b374f69 same as 1952 (was ca78893d)
GDD Phase 5 PR 3: recruit formations and their upkeep (#742, #1130) smoke5 d449bf19 same upkeep over the five smoke turns; no snapshot is kept for this run (was 71a4cef0)
GDD Phase 5 PR 3: recruit formations and their upkeep (#742, #1130) smoke-orders ee6bfee5 the same upkeep over its one turn; no snapshot is kept for this run (was 4c991630)
GDD Phase 5 PR 3: recruit formations and their upkeep (#742, #1130) scenario-reckless 6650ed3e same; debt alone rises by IL 3.4m (was 57b3c29d)
GDD Phase 5 PR 3: recruit formations and their upkeep (#742, #1130) scenario-cautious 95982cf8 same as reckless (was d6e7cea0)
GDD Phase 5 PR 3: recruit formations and their upkeep (#742, #1130) scenario-mvp-absorb-first d6d47fd8 same; debt alone rises by IL 8.2m (was 912504be)
GDD Phase 5 PR 3: recruit formations and their upkeep (#742, #1130) scenario-mvp-defeat 584a828a same; debt alone rises by IL 17.8m; the defeat is unchanged (was 109b0a5e)
GDD Phase 7 PR 6, slice 1: the score's parts (R-SCORE-001, #809) scenario-mvp-absorb-first 6aa49f62 the state keeps the score's parts as they score, a new scoreparts= line (objective points, timing bonuses, prisoner points): scoreparts=100,0,0 beside score=100. Nothing else moves; the other goldens score nothing, so they keep their hashes (was d6d47fd8)
#80: rationing, rationing_feeds and rationing_grates from 10 March 1949 (owner ruling 22 Oct 2026) smoke5 bc2ab42b the rationing policy and its two effects leave the state before 10 March 1949, so their keys (all at 0 in this run) leave the hashed state; no value moves (was d449bf19)
#80: rationing, rationing_feeds and rationing_grates from 10 March 1949 (owner ruling 22 Oct 2026) smoke-orders 9f224bd4 same as smoke5 (was ee6bfee5)
#80: rationing, rationing_feeds and rationing_grates from 10 March 1949 (owner ruling 22 Oct 2026) scenario-cautious f672a0f3 rationing can no longer be set on 15 May 1948, so the scenario sets it (0.4) on the first turn the office exists (15 March 1949) and runs to 15 April 1949 (was 15 January 1949) so it acts; the snapshot moves with the three extra turns (was 95982cf8)
#80: rationing, rationing_feeds and rationing_grates from 10 March 1949 (owner ruling 22 Oct 2026) scenario-reckless cb1c96d5 same as cautious, rationing 1.0 (was 6650ed3e)
#798: the mvp-occupy-cost golden (the plan's PR 6 scenario; owner, 23 Oct 2026, now that a capture is reachable in play, #1304) scenario-mvp-occupy-cost a767bb66 new: Golani's unit moves to Nazareth on 15 May 1948, the scenario answers the capture with Occupy on 15 June, and the run ends on 15 December 1948 with Nazareth awaiting EndOccupation (since=1948-06-15). No existing golden moves
#799: the mvp-annex-nazareth golden (the plan's Task 6, #1520) scenario-mvp-annex-nazareth d37e6db4 new: mvp-occupy-cost's two steps, then end_occupation nazareth annex on 15 December 1948, run to 15 January 1949. Nazareth ends owner=israel,state=annexed,since=1948-06-15 and is a city of 17 units, 7 employed (16,994 people, the census-day count the owner approved on #799, times the national share 0.439), housing 17. It pins the approved count. The fled field and EndOccupation move no existing golden: mvp-occupy-cost captures Nazareth (stayed) and keeps a767bb66
#1565: the degania_defence card on turn 1 (#1568; owner, 24 Oct 2026: "Keep +0.015", "Accept the shift") 1950 8578966e the reference player answers the new 15 May 1948 card with its historical jordan_valley option (security +0.015); answering two cards takes that turn's Attention, so the first halutz group is trained on 15 August 1948 instead of 15 May. Security, morale and stability rise a little and debt is IL 2.0m lower at 1950 (was f3ec68b7)
#1565: the degania_defence card on turn 1 (#1568; owner, 24 Oct 2026: "Keep +0.015", "Accept the shift") 1952 5bfeb265 the same card and later first training; by 1952 debt is back to the old run's value, while security, morale and stability stay a little higher. The last required yishuv (Re'im) is built on 15 September 1951, not 15 July, so first_new_yishuv fires two months later (was 6b374f69)
#1565: the degania_defence card on turn 1 (#1568; owner, 24 Oct 2026: "Keep +0.015", "Accept the shift") mvp-end 5bfeb265 same as 1952 (was 6b374f69)
#1565: the degania_defence card on turn 1 (#1568; owner, 24 Oct 2026: "Keep +0.015", "Accept the shift") smoke5 7bdf646a the first option of the new card over the five smoke turns; no snapshot is kept for this run (was bc2ab42b)
#1565: the degania_defence card on turn 1 (#1568; owner, 24 Oct 2026: "Keep +0.015", "Accept the shift") smoke-orders bf0b41cc same over its one turn; no snapshot is kept for this run (was 9f224bd4)
#1565: the degania_defence card on turn 1 (#1568; owner, 24 Oct 2026: "Keep +0.015", "Accept the shift") scenario-reckless a2ca9626 the scenario's player answers the card jordan_valley: security +0.015, with morale and stability following (was cb1c96d5)
#1565: the degania_defence card on turn 1 (#1568; owner, 24 Oct 2026: "Keep +0.015", "Accept the shift") scenario-cautious c06d89f0 same as reckless (was f672a0f3)
#1565: the degania_defence card on turn 1 (#1568; owner, 24 Oct 2026: "Keep +0.015", "Accept the shift") scenario-mvp-absorb-first 9c98ae51 same as reckless (was 6aa49f62)
#1565: the degania_defence card on turn 1 (#1568; owner, 24 Oct 2026: "Keep +0.015", "Accept the shift") scenario-mvp-defeat d5bc68f9 same as reckless; the defeat is unchanged (was 584a828a)
#1565: the degania_defence card on turn 1 (#1568; owner, 24 Oct 2026: "Keep +0.015", "Accept the shift") scenario-mvp-occupy-cost 9fcc7bae same as reckless; Nazareth's capture and Occupy are unchanged (was a767bb66)
#1565: the degania_defence card on turn 1 (#1568; owner, 24 Oct 2026: "Keep +0.015", "Accept the shift") scenario-mvp-annex-nazareth 8fe6d129 same as reckless; the annexation is unchanged (was d37e6db4)
#939 Task 12: accepted slice-one follow-up cards all manifest runs unchanged jerusalem_demilitarisation_guard and rafah_force_pressure fire only after their approved alternative roads. STATECRAFT_UPDATE_GOLDEN=1 dotnet test sim/Statecraft.sln -c Release on 24 October 2026 passed 2,600 tests; no golden moved. The content hash changes, so earlier saves retain the content-mismatch warning.
#1564: the reparations_vote card, the Knesset vote on talks with West Germany (fires 15 December 1951) hash1952 6dfd2337 the new hard card fires on turn 43 of the historical run, the reference player answers it historically (negotiate, morale -0.02), and the run's state and fired events change (was 6b374f69)
#1564: as above mvp-end 6dfd2337 same run as hash1952 (was 6b374f69)
#1564: as above scenario-mvp-defeat 85c86120 the scenario's run reaches December 1951, so the card fires in it too and, left unanswered, resolves to negotiate (morale -0.02; was 584a828a)
#1575 + #1568 combined: the reparations_vote card (#1564) on top of degania_defence (#1565) and #939 Task 12 (#1545) hash1952 385ba3b5 re-pinned on the merged tree: the reference player answers reparations_vote negotiate on 15 December 1951 (morale -0.02, stability -0.0015 through it), as on #1575's own branch; degania_defence's changes stay (was 5bfeb265)
#1575 + #1568 combined: as above mvp-end 385ba3b5 same run as hash1952 (was 5bfeb265)
#1575 + #1568 combined: as above scenario-mvp-defeat b7d3848d the card fires in the scenario's run too and, left unanswered, resolves to negotiate (morale -0.02; was d5bc68f9). No other golden moves: hash1950, smoke5, smoke-orders and the other five scenarios end before 15 December 1951
#1183: the 4.4 km hex bake (owner, 21 Oct 2026; one re-pin, alone in its merge slot) hash1950 74fbe989 every unit and the reference player's Tel Aviv housing on its 4.4 km hex (veteran_halutz/housing@tel_aviv,-3,19 becomes -3,21); debt IL 0.1m higher: the opening fights differ on the new grid (Golani's 15 May attack goes to the Iraqi unit next to it at (11, 12), the Syrian start anchor being two hexes away, ReferencePlayerTests). A behaviour change, not only coordinates: the 7th Brigade (unit 13) is cut to strength 1. Its 24 May anchor, Hulda, falls in hex (-4, 25), which now neighbours the Arab Legion's Latrun hex (-3, 25) (on 5 km Hulda's hex (-4, 23) was two hexes from the Legion's (-2, 22)). The enemy AI's Legion attacks it on 15 October 1948 (the Legion loses 1,000) and 15 November (the 7th loses 1,000 and retreats to (-4, 24), Rehovot's hex, where it stands to the end: 13:veteran_soldier@-4,24|side=israel|strength=1|full=2); on main hash1950 has 13:veteran_soldier@-4,23|side=israel|strength=2. Its retreat skips (-3, 24) because that hex is now Ramla's capture target (owner on #1796, 26 October 2026, "Contested + move target"; a retreat never ends on an enemy-held target, R-CTRL-001); before that move it retreated there and was destroyed on 15 December. It then takes (-4, 24), which Rehovot shares with Ramla's settlement: since the target move, Movement.NoEntry closes Ramla's target hex, not its settlement hex, for the player as Closed does (owner review of #1796, 31 October 2026); before that fix it went on to (-5, 25) and stood there at strength 1. No rule or value changed beyond that target (was 8578966e)
#1183: the 4.4 km hex bake (owner, 21 Oct 2026; one re-pin, alone in its merge slot) hash1952 d7429694 same run as hash1950 to the end, the 7th Brigade at strength 1 on (-4, 24) (see hash1950); debt IL 0.1m higher (was 385ba3b5)
#1183: the 4.4 km hex bake (owner, 21 Oct 2026; one re-pin, alone in its merge slot) smoke5 9255f2bc the first-option run on the new grid's positions (was 7bdf646a)
#1183: the 4.4 km hex bake (owner, 21 Oct 2026; one re-pin, alone in its merge slot) smoke-orders 08d9800e the one move goes to the first lit hex, (8, 12), which is no place at 4.4 km (at 5 km it skipped Nazareth's, SliceSmokeTests) (was bf0b41cc)
#1183: the 4.4 km hex bake (owner, 21 Oct 2026; one re-pin, alone in its merge slot) scenario-reckless 63cab812 unit positions on the 4.4 km grid; debt IL 0.05m lower. The 7th Brigade is lost to the Arab Legion (attacks on 15 October, 15 November and 15 December 1948: here it holds (-4, 25) after the second, where in hash1950 it retreats to (-4, 24) and survives), and here the Legion loses 1,000 in two of them, so it ends at strength 3, not 4; the owner accepted the loss with no special case (#1796, 27 October 2026, "Accept it") (was a2ca9626)
#1183: the 4.4 km hex bake (owner, 21 Oct 2026; one re-pin, alone in its merge slot) scenario-cautious e7ad49bf same as reckless, the 7th Brigade lost and the Legion at strength 3 included (was c06d89f0)
#1183: the 4.4 km hex bake (owner, 21 Oct 2026; one re-pin, alone in its merge slot) scenario-mvp-absorb-first b589c9dc its project hexes re-keyed to the new grid (each city's new hex plus the old offset; Rehovot's (-4, 20) offset lands on Rishon LeZion's hex, so (-3, 24)); debt IL 0.5m lower; as in reckless, the 7th Brigade is gone and the Legion ends at strength 3 (the snapshots) (was 9c98ae51)
#1183: the 4.4 km hex bake (owner, 21 Oct 2026; one re-pin, alone in its merge slot) mvp-end d7429694 same run as hash1952 (was 385ba3b5)
#1183: the 4.4 km hex bake (owner, 21 Oct 2026; one re-pin, alone in its merge slot) scenario-mvp-defeat 9e2afc15 unit positions on the 4.4 km grid; the run's late indicators differ (housing, morale, standing) with the different fights; as in reckless, the 7th Brigade is gone and the Legion ends at strength 3 (the snapshots) (was b7d3848d)
#1183: the 4.4 km hex bake (owner, 21 Oct 2026; one re-pin, alone in its merge slot) scenario-mvp-occupy-cost 0bc59565 Golani's unit is 5 movement points from Nazareth at 4.4 km, so the scenario moves it on 15 May to (9, 11) and onto Nazareth on 15 June, answers Occupy on 15 July and runs to 15 January 1949 (since=1948-07-15); as in reckless, the 7th Brigade is gone and the Legion ends at strength 3 (the snapshots) (was 9fcc7bae)
#1183: the 4.4 km hex bake (owner, 21 Oct 2026; one re-pin, alone in its merge slot) scenario-mvp-annex-nazareth b6df4284 mvp-occupy-cost's steps one turn later, then end_occupation nazareth annex on 15 January 1949, run to 15 February 1949; as in reckless, the 7th Brigade is gone and the Legion ends at strength 3 (the snapshots) (was 8fe6d129)
#749 part 2: the operation cards' unit orders (R-ENEMY-005; owner, 31 Oct 2026: "GDD 0.1 only", "Sourced anchors only"; owner's answer 1 on #2087: "Orders expire with the operation") hash1950 01d5dea9 the reference player answers ten_days central_front on 15 July 1948: Carmeli (unit 1) sets out from Haifa toward its Brosh anchor and reaches 8,8, in the Lower Galilee, that turn; its order ends with the anchor's window (19 July), so on 15 August it stops there. It answers operation_horev launch on 15 December 1948: Golani (unit 0) sets out from Yavne'el toward the Julis camp, goes round Nazareth (no capture), and on 15 March 1949, after its order ended with the Egypt armistice (24 February), stops at 3,14, between Zikhron Ya'akov and Hadera, short of Julis. No fighting and no capture; the final state differs from main's only in where the two brigades stand (was 8578966e)
#749 part 2: as above hash1952 ece2105d same orders as hash1950; Golani stays at 3,14 and Carmeli at 8,8 to the end (was 385ba3b5)
#749 part 2: as above mvp-end ece2105d same run as hash1952 (was 385ba3b5)
#749 part 2: as above smoke5 6da1203f the smoke run's first option of ten_days sends Carmeli toward its Brosh anchor; it stops at 8,8 on 15 August 1948. No snapshot is kept for this run (was 7bdf646a)
#749 part 2: as above scenario-reckless a6328f59 Horev's delegated order takes Golani round Nazareth; it stops at 3,14 on 15 March 1949 (was a2ca9626)
#749 part 2: as above scenario-cautious 725c9465 as reckless, and the delegated ten_days sends Carmeli to 8,8, where it stops on 15 August 1948 (was c06d89f0)
#749 part 2: as above scenario-mvp-absorb-first e9b657e7 both orders, as in cautious; no capture and no fighting (was 9c98ae51)
#749 part 2: as above scenario-mvp-defeat 6dc76d20 both orders, as in cautious; no fighting, so no extra RNG draws, and the delegated cards and the defeat are as on main (was b7d3848d)
#749 part 2: as above scenario-mvp-occupy-cost 2110fe92 Carmeli stops at 8,8 as in cautious; Golani holds Nazareth for the scenario and the run ends on 15 December 1948, before Horev's order; Nazareth's capture and Occupy are unchanged (was 9fcc7bae)
#749 part 2: as above scenario-mvp-annex-nazareth 12110c4b both orders: Golani leaves the annexed Nazareth for Julis and is on its way at 5,12 when the run ends on 15 January 1949, its order standing to 24 February; the annexation is unchanged (was 8fe6d129)
#2087, the owner's rulings of 1 November 2026 ("military operations must have unit orders"; Dani: "Kiryati to Ramle"; Latrun: Harel toward the fort): Kiryati to Ramle replaces Yiftach to Lydda, and Yoav, Hiram and Latrun's press_offensive give orders to new sourced order targets (R-ENEMY-005; owner, 2 November 2026, option 1: order_targets in map/units_1948.yaml, not anchors) hash1950 f4a7288f ten_days on 15 July 1948: Kiryati (unit 3) reaches Ramle's hex the same turn and takes Lydda and Ramle (their residents flee, as for every capture); no decision is taken, so on 15 August it withdraws to the next hex, -2,21, and both are under enemy control again. operation_yoav on 15 October: Givati (unit 4) reaches Huleiqat; the Egyptian unit next to it attacks twice, and on 15 November Givati, down to strength 1 of 3, falls back to -7,26; the Negev Brigade (unit 7) marches south and takes Beersheba on 15 December, then withdraws to -8,31 on 15 January 1949, and Beersheba is under enemy control again. operation_hiram on 15 November: Carmeli (unit 1) sets out from 8,8 for Manara; the ALA attacks it on 15 January 1949 (strength 2 to 1) and it reaches Manara on 15 February. The 7th Brigade's order (unit 13) is refused on 15 November, every run, because the ALA's unit holds al-Malikiyya; the turn's refusals say so. The combat draws shift the RNG, so the Arab Legion's attack on Harel on 15 October costs it nothing (strength 5, was 4). The reference player answers latrun_proposal with the vote, so Harel's order is never given (was 01d5dea9)
#2087, as above hash1952 452562ce same as hash1950; nothing moves after 15 February 1949 (was ece2105d)
#2087, as above mvp-end 452562ce same run as hash1952 (was ece2105d)
#2087, as above smoke5 54f201ca the smoke run's first option of ten_days sends Kiryati to Ramle, which it takes on 15 July 1948 (was 6da1203f)
#2087, as above scenario-reckless ad599eb9 no ten_days in this run. Givati is destroyed at Huleiqat by the Egyptian unit's second attack on 15 November 1948; the Negev Brigade takes Beersheba on 15 December and still holds it, awaiting a decision, when the run ends on 15 April 1949; Carmeli, sent from Haifa toward Manara, is destroyed by the ALA at 13,4 on 15 February 1949; the 7th Brigade's order is refused on 15 November (was a6328f59)
#2087, as above scenario-cautious d8724383 Kiryati takes Lydda and Ramle on 15 July 1948 and holds them, awaiting a decision, to the run's end; Givati is destroyed at Huleiqat on 15 November; the Negev Brigade takes Beersheba on 15 December; Carmeli is destroyed by the ALA on 15 January 1949; the 7th Brigade's order is refused (was 725c9465)
#2087, as above scenario-mvp-absorb-first a2b5fe92 as cautious, to 15 January 1950 (was e9b657e7)
#2087, as above scenario-mvp-defeat 918b56ff as cautious; the extra combat draws shift the RNG, so the delegated iraqi_exodus card draws airlift instead of pace (morale, international standing and housing differ from that card on), and on 15 December 1951 the last turn's withdrawal takes Kiryati out of Lydda and Ramle and the Negev Brigade out of Beersheba. The run still ends in defeat (was 6dc76d20)
#2087, as above scenario-mvp-occupy-cost 59cae466 the run ends on 15 December 1948: Kiryati holds Lydda and Ramle, awaiting a decision; Givati was destroyed at Huleiqat on 15 November; the Negev Brigade is one hex short of Beersheba and Carmeli on its way at 10,7, both orders standing; Nazareth's capture and Occupy are unchanged (was 2110fe92)
#2087, as above scenario-mvp-annex-nazareth 3f92f80e the run ends on 15 January 1949: Kiryati holds Lydda and Ramle and the Negev Brigade Beersheba, all awaiting a decision; Givati was destroyed at Huleiqat; Carmeli is on its way at 11,6; the annexation is unchanged (was 12110c4b)
#2087 merged with main after #1183 (the 4.4 km grid): the order targets and the cards' order hexes re-derived with tools/unit_anchor_hexes.py (Kiryati's Ramle, since e5e74883d Ramla's capture hex (-3, 24), Carmeli's Manara (15, 3), the 7th Brigade's al-Malikiyya (14, 4), Givati's Huleiqat (-9, 29), the Negev Brigade's Beersheba (-10, 35), Harel's Latrun (-3, 25); Golani's Julis anchor (-8, 28) and Carmeli's Brosh anchor (14, 6) are main's), one re-pin on the merged tree hash1950 48abe729 only the unit positions and one capture move; the economy lines are main's. The delegated and answered card orders now run on the 4.4 km hexes: Golani's Horev order (to 24 February 1949) leaves it at 4,14; Carmeli ends at Manara (15,3); Kiryati stops on (-4,24) when the Dani order ends on 20 July 1948 (the next rows run it to 16 August, and it takes Ramla); Givati holds Huleiqat (-9,29); the Negev Brigade takes Beersheba on 15 January 1949 and ends next to it at -9,35. The 7th Brigade's Hiram order to (14,4) is refused on 15 November 1948 while the ALA holds that hex, as before (was 74fbe989)
#2087 merged with main after #1183, as above hash1952 43061114 same run as hash1950; nothing moves after 1950 (was d7429694)
#2087 merged with main after #1183, as above mvp-end 43061114 same run as hash1952 (was d7429694)
#2087 merged with main after #1183, as above smoke5 a968abbc the smoke run's first option of ten_days sends Kiryati toward Ramle and Carmeli to (14,6). No snapshot is kept for this run (was 9255f2bc)
#2087 merged with main after #1183, as above scenario-reckless 4f3ca893 no ten_days in this run. Golani ends at 4,14 as in hash1950; Carmeli is on its way to Manara at 13,3, its order standing to 23 March 1949; Givati holds Huleiqat; the Negev Brigade takes Beersheba on 15 January 1949 and holds it, awaiting a decision (was 63cab812)
#2087 merged with main after #1183, as above scenario-cautious be999e93 as hash1950, with Beersheba awaiting a decision at the run's end (was e7ad49bf)
#2087 merged with main after #1183, as above scenario-mvp-absorb-first 1e14658d as cautious (was b589c9dc)
#2087 merged with main after #1183, as above scenario-mvp-defeat 8435fb3b the unit positions and Beersheba's capture as in hash1950; the defeat is unchanged (was 9e2afc15)
#2087 merged with main after #1183, as above scenario-mvp-occupy-cost f3bab7ca the run now stops on 15 December 1948, one turn earlier: the 15 December turn delegates Operation Horev, whose order would take Golani off Nazareth and end the occupation this scenario measures. Nazareth still waits for EndOccupation since 15 July 1948; the one-turn-shorter run differs in the economy, the events and the arrivals, and on 15 December the 7th Brigade still holds (-4,25) at strength 1 under the Legion's attack (was 0bc59565)
#2087 merged with main after #1183, as above scenario-mvp-annex-nazareth ca015736 Nazareth is annexed on 15 December 1948, a turn earlier, in the turn that delegates Horev, before its order takes Golani away (a month later there would be no occupation to end); the upkeep stops a month earlier (debt IL 0.04m lower). Golani is on its way toward Julis at 4,14 and Carmeli toward Manara at 15,5 when the run ends on 15 February 1949; Beersheba is taken on 15 January 1949 and awaits a decision (was b6df4284)
#2087, the owner's Dani ruling of 2 November 2026 ("Longer order: takes Ramla (Recommended)"): Kiryati's Ramle order target runs to 16 August 1948 instead of 20 July (map/units_1948.yaml), one re-pin on the tree merged with main 573b9d7ac hash1950 ac38d6c7 the 15 July 1948 turn still leaves Kiryati on (-4,24), one hex short of Ramla's capture hex (-3,24) after going round Lod; the order now stands on 15 August, so Kiryati steps onto (-3,24) and takes Ramla that turn (since=1948-09-15, fled). Lod is not taken by the card. The reference player then raids Ramla (HistoricalAI.cs:24-27), so Kiryati withdraws to -2,24 and Ramla is under enemy control again at the run's end (an ownerless record, fled); the other units, Beersheba and the economy lines are unchanged (was 48abe729)
#2087, the longer Dani order, as above hash1952 22f2c8da same run as hash1950: Ramla taken in the 15 August 1948 turn, then raided, so Kiryati is at -2,24 and Ramla is under enemy control at the end (was 43061114)
#2087, the longer Dani order, as above mvp-end 22f2c8da same run as hash1952 (was 43061114)
#2087, the longer Dani order, as above smoke5 803f0f32 the smoke run's first option of ten_days gives the longer Dani order, so its run to 15 October 1948 differs from the 15 August turn on. No snapshot is kept for this run (was a968abbc)
#2087, the longer Dani order, as above scenario-cautious 0be86235 Kiryati takes Ramla in the 15 August 1948 turn and holds it on (-3,24), Ramla awaiting a decision at the run's end (was be999e93)
#2087, the longer Dani order, as above scenario-mvp-absorb-first 53c208c2 as cautious (was 1e14658d)
#2087, the longer Dani order, as above scenario-mvp-defeat 6782cce4 Ramla taken in the 15 August 1948 turn, then raided, so Kiryati is at -2,24 and Ramla is under enemy control at the end, as hash1950; the defeat is unchanged (was 8435fb3b)
#2087, the longer Dani order, as above scenario-mvp-occupy-cost 612e9feb as cautious: Kiryati holds Ramla on (-3,24), awaiting a decision; the run still stops on 15 December 1948 (was f3bab7ca)
#2087, the longer Dani order, as above scenario-mvp-annex-nazareth c8195c09 as cautious: Kiryati holds Ramla on (-3,24), awaiting a decision; Nazareth's annexation is unchanged (was ca015736)
#1270 (S1, the Syrian script past the first truce; Beri's choice (a) on #2180, 3 Nov 2026, under the owner's 22 October rule), re-pinned on main after #2087 hash1950 a08ff850 the Syrian unit's script runs past 8 July 1948 to the bridgehead's anchor S1 (to 20 July 1949). On the 4.4 km grid S1's point falls in Rosh Pinna's hex (14, 6), a player's locality, and (15, 6) is already Syrian anchor 1, so tools/unit_anchor_hexes.py puts S1 on the free neighbour (15, 5): on the 15 July 1948 turn the unit steps one hex from its June hex (15, 6) onto (15, 5) and holds it fortified (10:syrian_army@15,5). With #2087's operation-card orders, unit 1's card go-to toward (15, 3) now ends at (15, 4), one hex short of (15, 3) (1:veteran_soldier@15,4) (was ac38d6c7)
#1270 (S1), re-pinned on main after #2087: as above hash1952 5561fc51 same run as hash1950: the Syrian unit on (15, 5), unit 1 on (15, 4) (was 22f2c8da)
#1270 (S1), re-pinned on main after #2087: as above mvp-end 5561fc51 same run as hash1952 (was 22f2c8da)
#1270 (S1), re-pinned on main after #2087: as above smoke5 d12c83ce the smoke run's five turns reach 15 October 1948, past the 15 July step; no snapshot is kept for this run (was 803f0f32)
#1270 (S1), re-pinned on main after #2087: as above scenario-reckless bed7eb74 the Syrian unit on (15, 5) is the only change (was 4f3ca893)
#1270 (S1), re-pinned on main after #2087: as above scenario-cautious 4f3a0556 the Syrian unit on (15, 5); unit 1 is still on its card order at the run's end, on (15, 4) toward (15, 3), so the run keeps cardorders{ 1=1949-03-23 } (was 0be86235)
#1270 (S1), re-pinned on main after #2087: as above scenario-mvp-absorb-first f289f897 the Syrian unit on (15, 5), unit 1 on (15, 4) (was 53c208c2)
#1270 (S1), re-pinned on main after #2087: as above scenario-mvp-defeat 8c83f6e0 as absorb-first (was 6782cce4)
#1270 (S1), re-pinned on main after #2087: as above scenario-mvp-occupy-cost ad9af425 the Syrian unit on (15, 5) is the only change (was 612e9feb)
#1270 (S1), re-pinned on main after #2087: as above scenario-mvp-annex-nazareth e8a9dbf1 the Syrian unit on (15, 5); unit 1, on its card order toward (15, 3), is on (13, 6) at the run's end instead of (15, 5) (was c8195c09). No other golden moves: smoke-orders does not reach 15 July 1948
#2608: the annexed residents join the population (R-CITY-008) scenario-mvp-annex-nazareth d0372ca6 end_occupation nazareth annex adds the row's count to GameState.Population's annexed group, so the population gains the annexed residents, and the debt, FX and Food lines move with it; the annexation itself and every unit position are unchanged. No other golden occupies a place, so none moves (was e8a9dbf1)
#2804 item 4: GameState.TotalPopulation sums the population groups in ordinal key order (R-FOOD-001; owner, 6 Nov 2026: "Do it just before the regen") hash1950 7d8e158b floating-point rounding only: the same numbers added in a different order land one ulp apart, in debt (...5F79 to ...5F7A); no rule, value or event changes (was a08ff850)
#2804 item 4: as above hash1952 f5ebb830 rounding only: fx moves by 16 ulp (-12.990431328096559 to -12.99043132809653) and nothing else (was 5561fc51)
#2804 item 4: as above mvp-end f5ebb830 same run as hash1952 (was 5561fc51)
#2804 item 4: as above smoke5 b93ad550 rounding only; no snapshot is kept for this run (was d12c83ce)
#2804 item 4: as above scenario-reckless c0eeb7dd debt 2 ulp apart, nothing else (was bed7eb74)
#2804 item 4: as above scenario-cautious 31ab04b0 debt 2 ulp apart, nothing else (was 4f3a0556)
#2804 item 4: as above scenario-mvp-absorb-first 16c8a869 debt 1 ulp apart, nothing else (was f289f897)
#2804 item 4: as above scenario-mvp-defeat 572f71ab debt and fx 2 ulp apart each; the defeat is unchanged (was 8c83f6e0)
#2804 item 4: as above scenario-mvp-occupy-cost eb3ae67b debt 1 ulp apart, nothing else (was ad9af425)
#2804 item 4: as above scenario-mvp-annex-nazareth 3cab078b debt 2 ulp apart, nothing else (was d0372ca6). No other golden moves: smoke-orders keeps 08d9800e
#2551, #2833 (R-TURN-002: ten decision cards a turn earlier), re-pinned on main after #2804 item 4 hash1950 d53edcd5 armistice_egypt comes on 15 February 1949 and signing it sets rung 2 a turn earlier, so the war costs one turn less (debt IL 3.75m lower); austerity_1949 is imposed on the 15 April 1949 turn, so rationing feeds one turn longer (food, morale and stability move). israel_bonds keeps its date (owner, 4 November 2026) (was 7d8e158b)
#2551, #2833: as above hash1952 25fbfb8f the same IL 3.75m lower debt and the drift of the earlier austerity; #2833's two cards (iraqi_exodus on 15 February 1951, forex_crisis_1951 on 15 October 1951) move morale, homes_lift_morale, fx (2 ulp) and part of stab: a run with only their two from lines reverted gives 4b7d3e79, where those lines are as before (was f5ebb830)
#2551, #2833: as above mvp-end 25fbfb8f the same run as hash1952 (was f5ebb830)
#2551, #2833: as above scenario-reckless ac5fbdd3 the run ends on 15 April 1949: Lebanon and Jordan now come on 15 March and are signed (international standing +0.10), austerity_1949 is pending at the end, and the debt is IL 3.75m lower (was c0eeb7dd)
#2551, #2833: as above scenario-cautious 46655872 as reckless (was 31ab04b0)
#2551, #2833: as above scenario-mvp-absorb-first cf73bb07 as hash1950 (was 16c8a869)
#2551, #2833: as above scenario-mvp-defeat 8096c41e its answers to austerity_1949 and reparations_claim move with the cards (15 April 1949, 15 February 1951); the debt is IL 3.75m lower and the run still ends in defeat; as in hash1952, #2833's two cards move morale, homes_lift_morale, both housing lines (rounding only) and part of stab (16ef2d5b with their from lines reverted) (was 572f71ab)
#2551, #2833: as above scenario-mvp-annex-nazareth dec2cad5 the run ends on 15 February 1949, the turn armistice_egypt now comes, so the card is fired and pending at the end (was 3cab078b). smoke5, smoke-orders and scenario-mvp-occupy-cost end before 15 February 1949 and do not move
#2543 operations need units in place (owner: "Player must move them"); the automated players stage the units like a player hash1950 b2dd7387 the reference player stages Kiryati, Carmeli, the 7th Brigade and Golani before their operation cards; all four operations still take their historical options and Rhodes and Sinai still fire; Ramla falls a month earlier (15 August), and one Golani-Egypt engagement replaces an old Legion one (same draw count) (was d53edcd5, main after #2551 and #2833; the same lines change as before them)
#2543 operations need units in place (owner: "Player must move them"); the automated players stage the units like a player hash1952 488d99fe carries hash1950's staged-operation state forward: same choices and follow-up cards, earlier Ramla capture, IL 0.1m less debt (was 25fbfb8f, main after #2551 and #2833; the same lines change as before them)
#2543 operations need units in place (owner: "Player must move them"); the automated players stage the units like a player smoke5 3803636f the five-turn first-option run gives no unit orders and stops at a first option whose operation units are not in place in the no-orders preview (the card shows that state), so Ten Days goes to its minister; with #2620's annexed residents on main (was b93ad550, main after #2551 and #2833; #2551 and #2833 do not move it, so the value is the same as before them; no snapshot is kept for this run)
#2543 operations need units in place (owner: "Player must move them"); the automated players stage the units like a player mvp-end 488d99fe the same 44-turn run as hash1952 (was 25fbfb8f, main after #2551 and #2833; the same lines change as before them)
#2543 operations need units in place (owner: "Player must move them"); the automated players stage the units like a player scenario-reckless 2217be01 Hiram, Yoav and Horev still resolve historically; staging leaves Golani in the south and the 7th Brigade alive; one engagement instead of two (one draw fewer) (was ac5fbdd3, main after #2551 and #2833; the same lines change as before them)
#2543 operations need units in place (owner: "Player must move them"); the automated players stage the units like a player scenario-cautious d421ba64 as scenario-reckless: one engagement instead of two; Carmeli reaches its Hiram target, the 7th Brigade survives, Ramla falls a turn earlier (was 46655872, main after #2551 and #2833; the same lines change as before them)
#2543 operations need units in place (owner: "Player must move them"); the automated players stage the units like a player scenario-mvp-absorb-first 5cded37c operations still historical; staging sends Golani to the Egyptian front (three engagements instead of two, one draw more) and the 7th Brigade survives (was cf73bb07, main after #2551 and #2833; the same lines change as before them)
#2543 operations need units in place (owner: "Player must move them"); the automated players stage the units like a player scenario-mvp-defeat 7fc8c624 as above, Ramla a turn earlier; one draw more, so the delegated iraqi_exodus card now takes pace instead of airlift (was 8096c41e, main after #2551 and #2833; the same lines change as before them)
#2543 operations need units in place (owner: "Player must move them"); the automated players stage the units like a player scenario-mvp-occupy-cost 90494f81 Ramla a turn earlier and Acre taken, the 7th Brigade survives, no combat draw; Golani stays Nazareth's last garrison (was eb3ae67b, main after #2551 and #2833; #2551 and #2833 do not move it, so the value is the same as before them)
#2543 operations need units in place (owner: "Player must move them"); the automated players stage the units like a player scenario-mvp-annex-nazareth adafb60f Golani is Nazareth's last garrison, so it is not staged and Horev postpones and lapses (the owner's rule): Rhodes Talks and Sinai Withdrawal never fire, and no combat draw (was dec2cad5, main after #2551 and #2833; the same lines change as before them)
#2543 review: the automated players' staging never targets an enemy-held place (Movement.NoEntry; R-ENEMY-007: staging is a move, not an attack) hash1950 832e8ccc Carmeli (unit 1) was staged for Ten Days onto Acre (8, 7), a Galilee member place, and captured it on 15 June 1948. It is now staged to another Galilee hex, so Acre is not taken and Carmeli ends a hex west, on (14, 3); nothing else differs (was b2dd7387)
#2543 review: the automated players' staging never targets an enemy-held place (Movement.NoEntry; R-ENEMY-007: staging is a move, not an attack): as above hash1952 841f0553 the same: no Acre capture, Carmeli on (14, 3) (was 488d99fe)
#2543 review: the automated players' staging never targets an enemy-held place (Movement.NoEntry; R-ENEMY-007: staging is a move, not an attack): as above mvp-end 841f0553 the same run as hash1952 (was 488d99fe)
#2543 review: the automated players' staging never targets an enemy-held place (Movement.NoEntry; R-ENEMY-007: staging is a move, not an attack): as above scenario-reckless 8bf27728 Acre is no longer captured and awaiting a choice at the end; nothing else differs (was 2217be01)
#2543 review: the automated players' staging never targets an enemy-held place (Movement.NoEntry; R-ENEMY-007: staging is a move, not an attack): as above scenario-cautious 024fa528 no Acre capture; Carmeli is on (14, 3) at the end, still on its card's go-to toward (15, 3), so the run keeps cardorders{ 1=1949-03-23 } (was d421ba64)
#2543 review: the automated players' staging never targets an enemy-held place (Movement.NoEntry; R-ENEMY-007: staging is a move, not an attack): as above scenario-mvp-absorb-first 96f62541 no Acre capture; Carmeli on (14, 3) (was 5cded37c)
#2543 review: the automated players' staging never targets an enemy-held place (Movement.NoEntry; R-ENEMY-007: staging is a move, not an attack): as above scenario-mvp-defeat 2c98e244 no Acre capture; Carmeli on (14, 3); the defeat is unchanged (was 7fc8c624)
#2543 review: the automated players' staging never targets an enemy-held place (Movement.NoEntry; R-ENEMY-007: staging is a move, not an attack): as above scenario-mvp-occupy-cost f9df2061 no Acre capture; Carmeli a hex west on its march, (10, 6) (was 90494f81)
#2543 review: the automated players' staging never targets an enemy-held place (Movement.NoEntry; R-ENEMY-007: staging is a move, not an attack): as above scenario-mvp-annex-nazareth cf8d2bdf no Acre capture; Carmeli a hex west on its march, (12, 4) (was adafb60f). smoke5 and smoke-orders do not move: the smoke runs stage nothing
#2543, the owner's 7 November 2026 rulings ("Wait, with a limit", "6 turns", "Keep the option"): a postponed operation waits for its units, up to mvp.operation_wait_turns turns, keeping its option, and the state records it (postponed{}, written only when set) smoke5 14066f39 the smoke gives no unit orders, so Ten Days, postponed on the 15 July 1948 turn, no longer lapses at its window's end: it is still pending and postponed on 15 October 1948, and the first-option run stops at it each turn, so Bernadotte and Yoav are pending behind it. No snapshot is kept for this run; the end state was read from a probe run (was 3803636f)
#2543, the owner's 7 November 2026 rulings ("Wait, with a limit", "6 turns", "Keep the option"): a postponed operation waits for its units, up to mvp.operation_wait_turns turns, keeping its option, and the state records it (postponed{}, written only when set): as above scenario-mvp-annex-nazareth 4d4a35d7 Golani is Nazareth's last garrison on 15 December 1948, so Horev is postponed rather than lapsing; once Nazareth is annexed the automated player stages Golani toward Horev's regions (read from HistoricalAI, which skips a last garrison), and on 15 February 1949, the run's end, Horev is still pending, postponed since turn 7, with Golani on (6, 12) on its go-to toward (-7, 26) (was cf8d2bdf). No other golden moves: in the other runs every operation is decided in its own window
#2543, the owner's Q4 of 7 November 2026 ("Shift by the delay"): postponed{} also records the date of the first postponement, from which a late launch's orders are shifted smoke5 12f088f6 format only: Ten Days' entry gains the date of its first postponement, turn 2, 15 July 1948 on monthly turns from 15 May; no snapshot is kept for this run (was 14066f39)
#2543, the owner's Q4 of 7 November 2026 ("Shift by the delay"): postponed{} also records the date of the first postponement, from which a late launch's orders are shifted: as above scenario-mvp-annex-nazareth c2aafd13 format only: operation_horev=launch@7 becomes operation_horev=launch@7@1948-12-15; nothing else differs (was 4d4a35d7)