Victory — Design Proposal

Every Last City · subsystem design · author: victory subsystem designer · clean-room

0. Glossary

Named systems, so the rest of this document can be terse.

1. The problem

Three problems, and they are not the same problem.

Problem A — victory is a checkbox. In games of this shape, "how you win" is usually a radio button with two or three entries, decided in ten seconds at setup and then forgotten. It has no presence during play. It generates no events. It is a terminating condition bolted onto a simulation, not a part of the game.

Problem B — the outcome is decided long before the game ends. This is the named structural flaw in §5 of the direction brief and the one that actually drives players away. It has three distinct sub-failures that people lump together:

• The mop-up. Everyone knows who won. You now spend an hour finding a broken opponent's last three units on a 2,000-tile map so the win condition can technically fire.
• The decided-but-not-over. Nobody has won yet, but the trajectory is obvious to all six players. Forty more turns of going through the motions.
• The stalemate. Two entrenched powers, neither able to crack the other, no clock. The game does not end at all; it is abandoned.

Each needs a different fix. A single "turn limit" setting fixes only the third and does it badly.

Problem C — you cannot see who is winning. Under fog, on a large map, a player often genuinely does not know they are losing until it is over. That is not tension, it is ignorance. Tension requires knowing the clock is running.

The design below treats victory as a system that runs all game, produces scheduled crises, publishes a legible scoreboard, and has four independent mechanisms for ending a game that is over.

2. The spine: everything is a Track, and Tenure is the currency

2.1 Every condition is one shape

Every victory condition in this design is a Track: an integer progress value, a threshold, and a Vigil. That uniformity buys three things that are worth more than the expressive freedom it costs:

1. One readout. Fourteen conditions, one bar widget. A player who understands the Dominion bar understands every bar.
2. One combination rule. "First Track to complete wins" is the whole of it.
3. One implementation. Adding a condition is adding a progress function and a threshold function, and the UI, the linter, the estimator and the AI evaluator all get it free.

Progress is always per-mille of threshold for the readout, so every bar in the game is comparable at a glance even when the underlying units are cities, regions, points and production.

2.2 The Vigil

A Track does not win the instant you cross it. It wins when you have held it across V consecutive Seals. This is the single most important mechanic in the proposal, and it is doing four jobs:

• It converts a threshold into a contest. You have to hold the thing, against everyone, while they all know exactly what you are holding and why.
• It creates the best line of text the game can produce: "Red completes Dominion in 3 turns." Nothing else in a strategy game generates that kind of pressure on demand.
• It kills flukes. A one-turn spike from a lucky double capture does not end a 180-turn game.
• It creates a deliberately asymmetric risk: the leader must be airtight for V turns; the challenger needs one successful attack. A five-turn investment can be undone by a single raid. That asymmetry is the anti-snowball mechanism, and unlike a hidden production tax it is one the player can see, plan for, and attack.

Failure resets the Vigil to zero, it does not decrement. Harsh on purpose. (Softer Decay variant exists as an option; see §11 tradeoffs — I am genuinely unsure which is right, and it is the first thing I would playtest both ways.)

2.3 Standing and Tenure

Standing is current score. It is deliberately short:

Worked, for Red at turn 178 of the example in §7 — 27 cities (2 of them Landmarks), his own Seat plus Gold's captured Seat, 132 output per turn, 38 units totalling 1,180 build cost:

cities     27 × 100                    = 2700
Landmarks   2 × 150                    = +300
Seats       2 × 200                    = +400
output    132 ×   5                    = +660
                                subtotal 4060
military  floor(1180 / 4) = 295, cap = floor(4060 × 25/100) = 1015 → +295
                                Standing 4355   → Tenure gain this turn = 43

Tenure = Σ over turns of floor(Standing / 100).

Tenure is where the design earns its keep, for two reasons.

Reason one — it is the universal tiebreaker. Every ambiguity in this document resolves to Tenure: simultaneous completion of different Tracks, the Horizon result, final placings for capitulated players, the Verdict sanity check. One rule instead of a precedence table, and it means the same thing every time: the player who has been ahead the longest.

Reason two — Tenure is monotone, so catch-up is exactly computable. Because Tenure only ever increases, and because the maximum a player can gain per turn is bounded (you cannot hold more than the whole map), the engine can compute exactly whether a trailing player is still able to reach the leader before the Horizon. That gives us a provable early-end trigger rather than a heuristic — the playoff-elimination arithmetic from professional sport, applied to a wargame. See §8.2.

Tenure is current-state integrated, never peak — you cannot bank a good turn 40 and coast.

2.4 Arithmetic discipline

All of it is integer. Per-mille everywhere, no percentages. Comparisons are cross-multiplied rather than divided (cities[L] × 2 ≥ cities[2nd] × 3, never a ratio). Drift is computed from the turn index by a closed form rather than accumulated per turn, so there is no accumulation error and no order dependence. Tenure is a floor-divided sum. Nothing here needs Q16.16, let alone a float.

3. The catalogue

Fourteen Tracks. Defaults given are for a medium map: 48 cities, 5 Landmarks, 8 regions, 4 players, Horizon 200 — the reference configuration used throughout this document.

The interesting ones in detail.

T1 — Dominion

Hold at least D(t) per-mille of all cities on the map for Vigil 3.

D is not a flat 60%. It scales with player count by a formula the player can say out loud: half the world, plus a share.

D_start = 500‰ + 500‰ / P

Progress is floor(1000 × cities_held / MAP_CITIES). MAP_CITIES is public from turn 1 even though city locations are not — you always know the size of the pie, never where the slices are. Without a public denominator the bar is meaningless, and hiding a scalar buys nothing.

The Ebb. From turn Ds (default 60% of the Horizon → turn 120), D falls on a published schedule to a floor of 501‰, reaching the floor exactly at the Horizon:

D(t) = D_start − floor( (D_start − 501) × (t − Ds) / (H − Ds) )     for t ≥ Ds

On the reference map: 625‰ at t=120, 587‰ at t=145, 536‰ at t=178, 501‰ at t=200.

The floor is 501‰ and not 500‰ so that at most one player can ever satisfy it, which removes a whole class of tie edge cases at the point where they are most likely.

The Ebb does three things. It guarantees termination. It ends stalemates by eventually bringing the bar down to whoever is nearest it. And — the part I care about most — because the schedule is published from turn 1, a player can count backwards to a date: "I hold 26 cities; the bar reaches me on turn 178; I need to survive 12 more turns." That is a scheduled endgame with a known day, which is precisely the thing a fog-of-war conquest game normally cannot produce.

Encourages: broad expansion, then broad defence — because losing two cities breaks a vigil. Breaks if: the map is small. On an 18-city map one city is 56‰ and vigils thrash. Linter rule L3.

T2 — Every Last City

Hold all of them. It is the purist option and it is the name of the game, which is a joke I am making on purpose: the title names the thing the design spends most of its effort refusing to make you do. It stays in, because some people want it and they are entitled to it, and because being honest about what it costs is better than pretending it does not exist. Its setup card carries a plain warning and force-enables Capitulation: Aggressive.

T4 — Hegemony

The map generator partitions the map into M named Regions, each with a named regional capital. Hold every city in K regions.

This is a genuinely different strategic shape from Dominion and that is why both exist. A player with 55% of the cities scattered opportunistically across every border has zero regions. Hegemony rewards clean borders and consolidation; Dominion rewards breadth. A setup that enables both asks the player to choose a doctrine.

Legibility is unusually good here: the map paints each region by completeness, so a region at 7-of-8 is visibly, achingly close, and everyone can see it.

T5 — The Keys

L Landmarks are seeded at generation (5 medium, 7 large — always odd). A Landmark is a city with a marker: better defensive terrain, higher Standing value, and a name. Hold K of L for Vigil 6. K = (L+1)/2, so 3 of 5 — already the uniqueness floor, so it does not Ebb.

Landmark locations are public to everyone from turn 1. This is a deliberate departure from fog purity and I will defend it directly: fog exists to hide what your opponent is doing, not what the world is. A general knows there is a mountain pass called the Iron Gate; he does not know who is marching to it. Publishing the Landmarks is what lets the game have known battlefields, and known battlefields are the single most effective anti-tedium device available — the war has an address before the war starts, and both sides can plan toward it from turn 1.

The long Vigil (6) is intentional. Keys should be fought over for a while, not sniped.

T6 — Ascendancy

Variant that solves fortress-squatting. At any time A of the L Landmarks are Lit (default 3 of 5), rotating every 12 turns. At each rotation there is a Tally: you score 1 Ascendancy point per Lit Landmark you hold at that Seal. First to 7 points wins.

The rotation schedule is derived from the game seed at generation and is published two rotations (24 turns) ahead. Everyone knows the Iron Gate lights on turn 96. Everyone can be there.

This is my favourite Track, because it manufactures scheduled crises on a metronome: sixteen tallies in a 200-turn game, each one a known date on which a known place matters. It rewards operational mobility over fortification, and it makes turtling actively lose.

T7 — The Great Work

A player designates a Work Site and pours production into it: 900 accumulated output. That is roughly 45 turns at a healthy 20/turn contribution.

• The Work Site must be a Landmark. Otherwise a player builds it in a far corner of a 2,000-tile map, nobody can reach it, and the game becomes an unreachable inevitability — which is exactly the failure mode we are here to eliminate. On a Landmark it is public, central, and defensible-but-attackable.
• Progress is visible to anyone who has scouted the site, and public to everyone at 50%.
• If the Site is captured, the captor gets nothing; the builder's stored progress is halved and may be re-sited after 5 turns. Total loss of a 45-turn investment is too swingy; zero loss makes capturing it pointless. Half is the compromise and I have no principled defence of the exact number.

Why this Track earns its place: it inverts the endgame. Instead of hunting for a hidden loser, everyone converges on a known place. The last twenty turns of a Great Work game are the best twenty turns in the design.

T9 — The Horizon

At turn H, highest Tenure wins. On by default in every preset; the linter shouts if you turn it off.

Resolving on Tenure rather than current Standing is deliberate and important. Tenure cannot be sniped on the last turn — you cannot buy the game on turn 199 with a desperate offensive, because the score you are being judged on was accumulated across 199 turns of actually holding things. It makes the readout a trend line rather than a snapshot, and it makes the final twenty turns about defending a record, which is a real strategic problem, instead of about a scoring spree, which is not.

Tie chain: Tenure → Standing → cities held → fewest units lost → lowest seat index.

T11 — Charters (per-player public objectives)

Asymmetric mode. Each player gets their own Track, visible to everyone. Red must reach Dominion 550‰; Blue must hold 4 of 7 Keys; Green must finish a Great Work; Gold must lead Tenure at turn 150.

The value here is information, not variety. Knowing that Green must go for Landmarks tells you where Green will be for the entire game. That is enormously richer than "everyone wants everything," and it gives a three-way game the texture of real strategy rather than free-for-all.

Charters are not equally hard, so each carries a Weight (1–5) and heavier Charters grant a setup handicap (an extra starting city, or a production head start). The weights are guesses. Say so; calibrate from self-play.

Two modes: Additional (shared Tracks also live) and Exclusive (only your Charter can win it for you). Exclusive is the more interesting game and the more fragile one.

T12 — Sealed Orders (secret objectives)

Secret victory conditions are usually a mistake, because being surprised by a loss is the worst experience a strategy game can deliver. Three rules make them safe:

1. A Sealed Order can never win the game by itself. It grants Ambition points (Minor 1, Major 2, Grand 3) toward a public track; first to 5 Ambition wins. Ambition is also obtainable publicly through Warrants, so the track is never purely hidden.
2. It auto-reveals at 60% progress. Nobody is ever surprised by a win. The reveal is the mechanic — it is the moment the table finds out what Green has been doing on that worthless island for thirty turns.
3. Their existence and difficulty class are public from turn 1. "Green holds two Sealed Orders, one Major." You know the shape of the threat without knowing its content.

Encourages: reading opponents' unexplained behaviour. Off by default in competitive presets.

T13 — Warrants (objectives that do not end the game)

The direction brief asks for objectives that create mid-game tension rather than only ending the game. Warrants are the direct answer, and they are the reason this is a system rather than a terminating check.

Every 15 turns the game publishes 1–3 Warrants, generated deterministically from the seed and announced 5 turns before they open, with a 15-turn deadline:

• "Hold Sable Reach at the Seal of turn 90: +200 Tenure."
• "Destroy 12 enemy units before turn 105: one city gains a production tier."
• "Be first to scout the Sunken Road: permanent vision of a 30-tile radius."
• "Hold both Lit Landmarks at the turn-96 Tally: +1 Ambition."

They inject scheduled conflict into the boring middle third. They pull players out of turtles, because an uncontested Warrant is free value and a contested one is a battle everyone chose to have. And they give the mid-game a rhythm — the flat stretch from turn 60 to turn 130 is the part of every long strategy game that sags, and this is the part of the design aimed squarely at it.

The rubber-banding question. Half the Warrants can be Contested: closed to the current Tenure leader. In-world that is the leader's rivals coordinating against them, which is at least an honest fiction. But this audience — Shadow Empire and Old World players — tends to resent being nannied. So it is an option: Warrants: Open / Contested / Mixed, Mixed in casual presets, Open in Tournament. I lean toward Open being right for the core audience and Mixed being right for the median buyer, and I do not think I can settle it without data.

T14 — Throne

Each player has a Seat. Losing it is severe but not fatal: you may re-Seat in another city after 6 turns, at a cost, and while Seatless you suffer a command penalty. The Track: hold every rival's Seat for Vigil 3.

Very few, very high-value, publicly-known targets. It is the fastest Track in the catalogue, which is what makes it the backbone of the short presets.

4. Combination rules

1. Conditions are a set, all live simultaneously, for all players (except Charters and Sealed Orders, which are per-player).
2. First Track to complete wins.
3. Completion is evaluated once per turn, at the Seal — after the entire activation queue has been walked, never mid-queue. This matters a great deal given the pipelined activation model in §4 of the direction brief: if victory were checked mid-queue, a player could win because their unit had a lower creation sequence number, and a turn model chosen to reduce waiting would have silently become a turn model that decides games. Evaluating at the Seal makes victory turn-atomic and independent of activation order. A city captured and recaptured within one turn nets out; only the Seal state counts.
4. Simultaneous completion of different Tracks resolves to higher Tenure, then the standard tie chain. One rule, no precedence table, and it means something defensible: the player who has been ahead longer.
5. Simultaneous completion of the same Track resolves to higher raw progress, then Tenure.
6. Vigils are independent and concurrent. Two players can both sit at 2-of-3 on Dominion. Only one can satisfy Dominion at the Ebb floor (501‰), by construction.
7. Vigil break resets to zero.
8. Only Sealed Orders may be hidden, and only under the reveal rules in T12.
9. Teams aggregate holdings and share Tracks and Tenure; a team wins as a unit.
10. Last Standing (T3) cannot be disabled. If one player remains, they have won. This is the only non-optional Track.

5. Legibility, part one: the Terms (setup)

A wall of fourteen conditions with parameters is its own failure. The setup screen is designed around three ideas.

Cards, not checkboxes. Each Track is a card: name, one plain-English sentence, at most three parameters, and three icons — Length, Aggression, Complexity — on a 1–5 scale. You should be able to read a card in four seconds.

A generated Terms paragraph. Below the cards, in prose, the game states what you just built:

This game ends when a player holds 30 of the map's 48 cities for 3 turns; or holds 3 of the 5 Landmarks for 6 turns; or at turn 200, when the highest Tenure wins. From turn 120 the Dominion requirement falls to 25 cities. Broken players surrender automatically after 3 turns. Estimated length: 90–160 turns.

That paragraph is the single most valuable object on the screen. If a player reads nothing else, they know what they signed up for.

Estimated length is real, not vibes. It comes from a shipped calibration table — median_turns[track][map_class][players] plus 20th/80th percentiles — generated at build time from ≥2,000 headless AI self-play games per cell. The deterministic headless core makes this cheap and the CI pipeline can regenerate it when balance changes. It is an integer lookup at runtime.

5.1 The Setup Linter

The combination rules in §4 are simple; the consequences of combining conditions carelessly are not. So the setup screen lints them, with concrete, numeric messages.

L9 is the one I most want to exist and am least sure I can compute honestly — see §12.

6. Legibility, part two: the Watch, the Board, the Alarm

6.1 The fog problem, and the answer

There is a real conflict: if you show everyone's exact progress, the scoreboard becomes a free intelligence oracle. "Red went 458‰ → 479‰ this turn, that's one city, so their army is in the north." You have handed away fog through the back door.

The answer is quantization that tightens as the stakes rise. Progress is public but its precision depends on how close the player is to winning:

This is the compromise done properly rather than split down the middle. Early, when information would leak troop movements, it is coarse. Late, when a player must be able to react, it is exact and loud. And there is an in-world reading that is not embarrassing: the world notices when someone is about to win.

Option Progress Visibility: Banded / Public / Own Only, Banded by default, Public in Tournament.

6.2 The Alarm and the World Watches

When a player enters Alarm, three things happen:

1. A full-width banner with a countdown: "RED HOLDS THREE KEYS — VICTORY IN 4 TURNS."
2. The contributing sites are named and listed, clickable to jump the camera.
3. Every player is granted live vision of those sites, including their garrisons, for the duration of the Alarm plus a 2-turn grace period after it clears.

Rule 3 is a large fog concession and it is the correct one. The failure being prevented is precise: the leader is about to win, I want to stop them, and I cannot find the thing I need to attack. A hidden win condition on a 2,000-tile map is not tension, it is a formality. The Alarm hands every player the address of the fight. That is what makes the last twenty turns a battle rather than a search — and a leader who reaches Alarm has, by design, accepted that they will be fought at a known place.

6.3 The Watch — always on screen

One compact strip, roughly forty pixels, never dismissed. One row per player, showing the Track they are nearest to completing:

 ● RED     ████████████░░░  Dominion    562‰/536‰   VIGIL 2/3   Tenure  15,140 ▲
 ● BLUE    ███████░░░░░░░░  Keys 2/3                            Tenure  11,410 ▲
 ● GREEN   ████░░░░░░░░░░░  Dominion    modest                  Tenure   6,020 ▼
 ○ GOLD    capitulated t141                                     Tenure   3,190 —
                                        Ebb: 536‰ → 501‰ by t200 · Horizon t180/200

Four facts per player, no clicking. Who is closest to what, by how much, and whether they are rising or falling.

6.4 The Board — one keypress

The full matrix: players × Tracks, per-mille of threshold, per-turn delta, Vigil state. Plus:

• The Ebb schedule as a descending line with a marker showing the turn each player's current holdings would suffice — the "count backwards to a date" readout.
• The Warrant board: open, upcoming, expired, and who claimed what.
• The Tally calendar for Ascendancy: which Landmarks light on which turns, 24 turns ahead.
• Charters and Sealed Order counts per player.
• The Chronicle: one line per victory-relevant event, permanently.

6.5 The Chronicle

 t131  Red takes Kestrel Bay (Landmark).  Red → Keys 3/3.  VIGIL BEGINS (6).  ALARM.
 t134  Blue takes Kestrel Bay.            Red → Keys 2/3.  VIGIL BROKEN at 3/6.
 t141  Gold capitulates (Broken 3 turns). 3 cities ceded to Red — Unsettled 8 turns.
 t145  Ebb: Dominion 587‰ (29 cities).
 t172  Horizon track DECIDED for Red — Blue and Green can no longer reach 14,880.
 t178  Ebb: Dominion 536‰ (26 cities).  Red holds 27.  VIGIL BEGINS (3).  ALARM.
 t179  Blue takes Ashford.  Red → 26 cities, 541‰ ≥ 536‰.  Vigil holds 2/3.
 t180  Vigil 3/3.  RED WINS — Dominion.

Anyone can reconstruct the whole endgame from twelve lines. That is the legibility target.

7. Worked example: one endgame, every mechanism

Reference map. 4 players. 48 cities, 5 Landmarks, 8 regions, Horizon 200. Tracks: Dominion (625‰, Vigil 3), Keys (3 of 5, Vigil 6), Horizon (Tenure). Capitulation: Standard (C=3). Verdict: on.

t=118. Red 22 cities, Blue 14, Green 8, Gold 4. Dominion progress: Red floor(22000/48) = 458‰ of a 625‰ bar = 73% → Report. Blue 291‰ = 47% → Rumour. Green 166‰, Gold 83‰ → Rumour. Keys: Red 2, Blue 2, Green 1. No Vigils. Tenure: Red 9,140 · Blue 7,020 · Green 4,880 · Gold 2,600.

t=120. The Ebb begins. The Board draws the descending line. Red reads it and gets a date: he holds 22, the floor is 501‰ = 25 cities, so 22 will never suffice — he needs 25 by turn 200, or 30 today. Blue reads the same line and learns he has until roughly turn 178 before 26 cities is enough for Red.

t=131. Red takes Kestrel Bay — his third Key. Keys 3/3, Vigil 6 begins. ALARM. Everyone gets the banner, the three site names, and live vision of Kestrel Bay, the Iron Gate and Sable Reach with garrisons. Blue can now see that Kestrel Bay holds four units and is nine tiles from his staging port.

t=134. Blue lands and takes Kestrel Bay. Red → Keys 2/3. Vigil resets from 3 to 0. Alarm clears; grace vision expires t=136.

This is the design working. A public six-turn countdown created a findable single point of failure, and one successful attack undid three turns of it. Under a classic instant-threshold rule Red simply wins at t=131 and nobody sees it coming.

t=138–141. Gold collapses. Gold falls to 1 city and 15% of peak output at t=138 → Broken. Broken at t=139, t=140, t=141 → capitulates at the Seal of t=141. Gold's units disband. Principal Aggressor: Red has taken 7 cities from Gold across the game, Blue 2 → Gold's remaining 3 cities cede to Red, arriving Unsettled for 8 turns (no output, no production). Gold's Tenure freezes at 3,190 and Gold takes a final placing.

Nobody spent forty turns hunting Gold's last destroyer. Red 26 cities.

t=160. Verdict check — and it correctly fails. Standing: Red 4,180 · Blue 2,510. 4180 ≥ 2 × 2510 = 5020? No — 1.67×, short of 2×. No Verdict. And rightly so: Blue still holds 2 Keys and is one capture from his own Vigil. The trigger is deliberately strict, and it should be visibly capable of declining.

t=172. The Horizon track is mathematically decided. Smax for this map = 4800 (cities) + 750 (Landmarks) + 800 (Seats) + 1200 (max output ×5) = 7,550, plus the military cap floor(7550 × 25/100) = 1,887 → 9,437. Max Tenure gain per turn = floor(9437/100) = 94. Remaining turns R = 28 → ceiling 2,632. Tenure: Red 14,880 · Blue 11,410 · Green 6,020. Blue: 11,410 + 2,632 = 14,042 < 14,880 → eliminated from the Horizon. Green: 8,652 < 14,880 → eliminated.

The Board marks the Horizon track DECIDED — RED. The game does not end: Dominion and Keys are still live and Blue can still win either outright. This is the honest limit of the mathematics and §8.2 says so plainly.

t=178. The Ebb reaches Red. D(178) = 625 − floor(124 × 58 / 80) = 625 − 89 = 536‰ = 26 cities. Red holds 27 (562‰). Dominion Vigil begins (3). ALARM. Live vision of Red's holdings that constitute the threshold.

t=179. Blue's counterattack takes Ashford. Red → 26 cities = 541‰. D(179) = 534‰. 541 ≥ 534 — the Vigil survives at 2/3. One city was not enough this time; Blue needed two.

t=180. D(180) = 532‰, Red 541‰. Vigil 3/3. RED WINS — Dominion, turn 180.

What the classic rule would have done. Total conquest only: at t=118 Red holds 22 of 48 and the game runs past turn 260 while Red ferries stacks to Blue's islands and Green's tundra. This design ended it at t=180 with a contested, visible, three-turn countdown that Blue nearly broke twice.

8. The endgame problem, answered

Four independent mechanisms, each aimed at a different sub-failure from §1.

8.1 Capitulation — kills the mop-up

A player is Broken at a Seal if either:

• cities_held = 0, or
• cities_held ≤ max(1, floor(peak_cities / 8)) and output ≤ 200‰ of peak output.

Continuously Broken for C turns → capitulation at that Seal.

On capitulation: units disband (no combat, no chase); remaining cities cede to the Principal Aggressor — the player who has captured the most cities from them across the whole game, tie broken by nearest city, then lowest seat index. Ceded cities arrive Unsettled for 8 turns (no output, no production), so the payout is not an instant snowball. If there is no aggressor, cities become garrisoned Free Cities. The player's Tenure freezes and they take a final placing.

Using whole-game captures rather than "who took the last city" is deliberate: it stops a third party swooping the final city to steal a war they did not fight.

8.2 Mathematical elimination — kills decided-but-not-over, provably

For accumulative Tracks — Tenure/Horizon, Ascendancy points, Warrant totals — the remaining ceiling is computable exactly:

ceiling(B) = remaining_turns × max_gain_per_turn
B is eliminated from track X iff  progress[B] + ceiling(B) < progress[leader]

The bound is deliberately loose (it assumes B takes the entire map while the leader gains nothing further), which makes it conservative and never wrong. When every player but one is eliminated on every enabled Track, the game ends immediately and the leader wins.

Honest limitation, stated up front: this does not work for instantaneous-threshold Tracks. There is no sound way to prove that a player with three cities cannot take thirty more in the remaining turns — in principle they can. So Dominion, Keys, Hegemony and Throne can never be closed by arithmetic, only by the Ebb and the Verdict. Any design claiming otherwise is smuggling in a heuristic and calling it a proof. Mine does not.

8.3 The Verdict — a vote, for the cases arithmetic cannot close

At a Seal, all four must hold:

1. Standing[leader] ≥ 2 × Standing[second]
2. The leader also leads on Tenure
3. The leader has held the Standing lead continuously for ≥ 12 turns
4. No non-leader is currently in any Vigil

Then the game offers a Verdict. Each surviving non-leader votes at their next turn start. Abstention counts as "play on" — the default is always to continue, because a wrongly ended game is far worse than a slightly long one. A majority of surviving non-leaders ends the game with the leader as winner and remaining placings by Tenure. A failed vote imposes a 20-turn cooldown.

Keying on Standing rather than Tenure is deliberate: Standing is current power and predicts the future, while Tenure integrates from turn 1 and moves too slowly for a 2× gap to appear before the game is long over. Tenure is the sanity check, not the trigger.

In solo play this is where most of the value lands. The AIs' votes are deterministic functions of their evaluation and the seed; a beaten AI concedes when it should, and the single-player mop-up — the worst version of the problem — largely disappears.

8.4 The Ebb and the Horizon — kill the stalemate, and guarantee termination

Threshold Tracks drift down on a published schedule to their uniqueness floor (501‰ for Dominion; >M/2 regions for Hegemony; already-floored for 3-of-5 Keys). Two entrenched powers cannot hold each other off forever, because the bar comes down to whichever of them is nearer it, on a date they both knew about a hundred turns in advance.

And the Horizon is a hard backstop with a defined, non-arbitrary resolution. The game provably terminates in at most H turns, always, with a winner determined by a total order (Tenure, then the tie chain). No abandoned games.

8.5 Concession

Always allowed; identical effects to capitulation. The conceder does not choose the beneficiary — the Principal Aggressor rule applies. This does not eliminate kingmaking in a three-player game (nothing can), but it makes the beneficiary whoever actually fought them, which is the fairest available answer, and it removes the ability to concede at someone out of spite. A conceder keeps their frozen Tenure and their placing, so playing on hopelessly has no scoreboard value either.

9. Objectives that are not endings

Worth stating as its own claim, because it is the part most easily lost in implementation:

A victory system should produce events every fifteen turns, not one event at the end.

Five mechanisms in this design fire during play and change what players do without ending anything:

The turn-60-to-turn-130 sag is the worst stretch of every long strategy game. Roughly nine Warrants and six Tallies land in it.

10. Presets

Every preset except Every Last City and Tournament runs the Verdict and the Ebb. A player who never opens the Terms screen gets Standard War, and Standard War is a good game.

11. Why this beats the classic approach

• Victory is present during play, not only at the end. Warrants, Tallies, the Ebb schedule and the Alarm produce roughly fifteen scheduled events in a 200-turn game. The classic version produces one, at the end, and it is the end.
• The Vigil turns a threshold into a fight. Crossing 60% is an announcement, not a victory, and one raid can undo five turns of it. That is a visible, attackable anti-snowball brake — exactly what §6 of the direction brief asks for in place of an invisible production tax.
• Banded visibility resolves fog-versus-legibility instead of picking a side. Coarse early so the scoreboard is not a troop-movement oracle; exact and loud late, when the information is what makes a counter-attack possible at all.
• The mop-up is removed rather than shortened. Capitulation deletes broken players from the board with a defined cession rule; no thresholds require 100% except the one option explicitly labelled as doing so.
• A provable early end exists for accumulative Tracks, with the limits of that proof stated honestly rather than papered over with a heuristic wearing a proof's clothes.
• Termination is guaranteed. The Ebb plus the Horizon means no game runs forever and none ends arbitrarily.
• One currency, one tiebreaker. Tenure resolves simultaneity, the Horizon, placings and the Verdict sanity check. There is no precedence table to memorise.
• Turn-one readability. City count, Landmark locations, region layout, Ebb schedule, Tally calendar and the first Warrants are all public before the first unit moves. A systems player can plan the whole game on turn 1, which is the thing this audience actually shows up for.

12. Tradeoffs and honest costs

• Public progress is information that did not exist before. Banding limits the leak but does not eliminate it, and some fog purists will feel the mystery deflate. I think the trade is correct, but it is a real cost and the Own Only option exists for people who disagree.
• Vigil reset is brutal. Losing a five-turn investment to one raid may read as unfair rather than tense. The Decay (−1) variant exists. I do not know which is right and I would run both.
• Tenure structurally favours the leader. Integrating score means an early lead compounds in the readout, and a comeback must be territorial rather than scoreboard-based. That is intentional — it is what makes the trend line mean anything — but it means Tenure is a bad comeback surface, and Warrants are carrying a load they may not be strong enough for.
• Rubber-banded Warrants will annoy the target audience. Contested Warrants are a concession to the median player and a provocation to the core one. Defaults split the difference, which is what you do when you cannot decide.
• Setup is genuinely heavier. Fourteen Tracks with parameters is more than three checkboxes. Presets and the Terms paragraph mitigate; they do not eliminate.
• Capitulation cession can snowball. Unsettled-for-8 is a guess. In a three-player game a capitulation may effectively decide the match, and the Principal Aggressor rule makes that fair without making it close.
• Standing couples victory to the economy subsystem. If output scaling changes, the ×5 weight and the 25% military cap need recalibration or Prosperity and Dominion collapse into each other. That is a permanent maintenance cost and it needs a CI check, not good intentions.
• The calibration table goes stale. Estimated lengths, linter rules L2 and L9, and AI Verdict policy all depend on self-play data that is invalidated by any balance change. If regenerating it is not automated, the estimates will silently become lies.
• L9 may not be computable honestly. Detecting that two Tracks are redundant requires a correlation measure over calibration runs, and correlation across map classes may be too noisy to threshold. It might have to become advisory prose on the cards instead of a lint.

13. What could go wrong, and how playtesting would catch it

The deterministic core is the strongest testing asset this design has, because most of these checks are counterfactual replays: re-run the exact game with one rule disabled and compare. That is unusual and I want to lean on it hard.

14. Open questions I could not settle

1. Vigil reset or decay? Reset is more dramatic and more punishing. I lean reset; I have no data and the argument for decay is decent.
2. Should Standing count military at all? Excluding it makes horizon-eve suicide attacks free. Including it rewards hoarding. The ÷4-capped-at-25% compromise is a guess, and the cap in particular is unfounded.
3. Is a hidden Horizon better? "The war ends somewhere between turn 180 and 220, exact turn unknown" would prevent last-turn optimisation entirely and is genuinely tempting. It is also much less legible, and legibility is the other half of the brief. Unresolved.
4. Prosperity may be a shadow of Dominion. Whether it is a distinct strategic shape depends entirely on how strongly the economy subsystem couples output to supply and development. If a sprawling 60%-of-cities empire can sit at 40% of potential output, they are different games. If not, Prosperity should be cut.
5. Does the Alarm's vision reveal hand a coalition a free strike? In a six-player game the leader entering Alarm may be jumped by five opponents with perfect information. That might be exactly right (winning should be hard) or it might mean nobody ever dares cross 90%, which would be a disaster.
6. Cede or free? Capitulation cession is clearly correct in a duel. In a five-player game handing three cities to the Principal Aggressor may be the largest single swing in the match. Free Cities may be better above four players.
7. Charter Weights and handicaps are pure invention. There is no defensible number here before self-play data exists.
8. Ambition needs its own numbers. First-to-5 with Minor/Major/Grand worth 1/2/3 is a sketch, not a design.
9. Teams and Tenure. Whether a team's Tenure is the sum of members' or the sum of the team's aggregate holdings changes what a sacrificial team member is worth. Not settled.
10. How does the Verdict work in async cloud play, where a vote may take days? Probably a default-to-"play on" timeout, but that makes the Verdict nearly useless in the mode where long games hurt most.

15. Contracts — what this subsystem needs from others