Decision: the turn model
Decided: Orders and the Cascade (see turn-model.md), over interactive activation on a
creation-order queue. Recorded here because it is the decision every other system keys off, and
because the reasoning matters more than the verdict.
The decision
Each turn has two phases:
- Orders — every player issues and revises orders concurrently, against a stable board, blocked by nobody. Bounded by a deadline in live play; a deadline measured in days in correspondence play.
- The Cascade — no input. The engine walks every unit in initiative order and resolves each order against the world as it stands at that instant. Computed in milliseconds, played back to each player as a fog-filtered animation.
Activation order is an integer Initiative score recomputed each turn, with creation sequence as the tiebreak — which is what guarantees the order is total, and therefore what makes the whole model paradox-free.
What decided it
Three arguments, in the order of how much weight they carried.
1. Correspondence play. Async is a first-class mode for this game — the play-by-email heir, and for a slow strategy game with scattered players it may be the mode that actually gets used. Under interactive activation, a turn requires one network round-trip per unit activation: six players with thirty units each is 180 sequential round-trips, potentially hours apart. That is not a slow correspondence game, it is a broken one. Under Orders and the Cascade it is exactly one submission per player per turn, whenever each player happens to sit down. The model does not merely tolerate async play; it is the natural shape for it.
2. You cannot run an operation under creation order. This was the design's sharpest point and it survives scrutiny. Empire-shaped play is about operations — these six units take the ford while the artillery displaces forward. Under a creation-order queue you do not choose the order your own units act in, and enemy actions land between them. You are reduced to thirty locally-sensible moves. Crucially, the interactive variant does not fix this — acting interactively when unit #7 comes up still means you cannot act with #9 first. The problem is the fixed ordering, not the interactivity.
3. It scales. Interactive activation costs one cold start per unit — someone has to notice their turn, orient, and decide. At eight players and 150 units each that is hours per turn. Concurrent ordering costs the maximum of the players' thinking time rather than the sum.
There is a fourth reason I weight quietly but genuinely: the command layer designed independently in
command.md — six Postures, four Sanctions, a three-deep order stack — is exactly the conditional
vocabulary this turn model needs to avoid degenerating into blind simultaneous orders. Two designers
working separately produced pieces that fit. That is usually a sign the shape is right.
What this costs, honestly
You commit before you see. This is a real change from the interactive feel of moving a unit and watching the result, and some players dislike it intensely. Three things mitigate it and none of them erase it:
- Orders are conditional and evaluated at activation, so a late-initiative unit's "advance if the ford is clear" sees everything that happened earlier in the Cascade. This is meaningfully less blind than pure WEGO.
- The model has strong company: Diplomacy, Combat Mission, and Frozen Synapse all commit before resolving, and in those games the commitment is the source of the tension rather than a tax on it.
- The Cascade playback is its own reward — watching a plan execute is a different pleasure from poking a unit, not a lesser one.
The conditional-order vocabulary becomes load-bearing. If it ships thin, the model degenerates into blind WEGO and the decision was wrong. This is the single biggest execution risk in the design and it must be treated as such: the Posture and Sanction vocabulary is not a convenience feature, it is the thing that makes committing early a skill rather than a gamble.
Amendments to the proposal
Two changes to what turn-model.md specifies:
Protect the small game. In solo and two-player games the Cascade MUST be able to run instantly with optional step-through, and the Orders phase MUST have no deadline unless one is asked for. Most play of a game like this is one person against AI opponents, where waiting is not a problem and the tactile pleasure of seeing each result is worth preserving. The scaling argument above is about the eight-player case; the design should not tax the common case to solve the rare one.
Trim initiative. Nine modifiers is too many to introduce at once and it fights the escalating-complexity goal. Ship with class base, posture, and Forming Up — three terms a player can hold in their head — and gate veterancy, command, supply, and disruption behind the same progressive disclosure the rest of the design uses. The tiebreak stays creation sequence throughout, because that is what keeps the order total.
What survives from the original idea
The load-bearing insight — that a single global total order over activations gives determinism and paradox-freedom for free, with no special cases for two units contesting a tile — is preserved exactly. It is the reason this model needs no simultaneous-move conflict rules at all. Creation sequence remains in the design as the tiebreak that makes the order total.
The seniority intuition also survives, inverted so it can be earned rather than merely accrued: new units carry a temporary Forming Up penalty that expires after four turns, and lasting initiative advantage comes from veterancy won in combat.