ADR-0016 — League metabolism: the game trails reality by one month

Status: ACCEPTED (shape; pipeline build + league-notes format owed) · Decided: 2026-07-13

These are the actual decision records the game is built from, published unedited except for internal process notes. They describe why the game works the way it does. They are not a strategy guide, and some of them argue with each other.

Status: ACCEPTED (shape; pipeline build + league-notes format owed)
Date: 2026-07-13

Context

The scouting beat's variance question ("what does a veteran re-scout per run?") was answered by moving the variance out of the seed and into time. Pip: "i'm considering making the leagues and associated patches that long as well - that means I can run the game a month behind real time, update real world events into the engine, balance patch, test, and deploy in a way that doesn't demand my full-time attention."

Pip's mature-cycle articulation: "collect real world events, suggestions on papers etc, value shifts from previous month's run of data, patch the game once a month and load the event logs into the game, the game's 'real time turn' progresses forward a month, a new seed is generated for seed-racers, and the league launches fresh with notes etc."

Decision

Monthly league cycle. Each real month: collect real-world events/papers/data shifts → author a world-update pack (ADR-0005 schedule entries, fed from the pdoom-data repositories) → the game's real-time frontier advances one month → new baseline seed(s) generated → league launches with notes.
Run-start stays 2017 (ADR-0009 anchor holds; Pip ~80% on this fork). Fixed real history is the shared macro map — Factorio-mode openings accepted at the macro grain; per-seed variance lives in the micro-social layer (hire pool, contacts, meetup yields). SC2 model: known map pool, scout the opponent's build; the pool rotates monthly. Parked experiment (the other ~20%): a "jump in at present-month" mode that runs 2017→present as baseline.
Cadences decoupled: world-updates ≠ balance patches. World-updates are monthly, light, expected — a "mild lore and sprint refresh." Balance patches are slower and legible-event-grade (philosophy canon: patches are the community heartbeat, argued about, never ambient weather). Real material is always welcome into pdoom-data; mechanics impact can lag at its own speed.
One month per baseline seed is the explore-exploit window — allows a few full playthroughs of a 6–8 hr decent run before rotation. Pip: timing "feels about right."
The ops constraint is a design input. The pipeline must run at ≤1 day/week of founder effort sustained: LLM-drafted world-update packs reviewed and submitted to Pip; junior-dev offload path if funding lands. Pip: "this can't take up more than 1 day a week of effort over long periods of time without it going public and actually having engagement pretty darn soon."

Beacons served / violated

Rams #6 (honest): the game is tethered to the real race — each league, another month of the actual timeline enters the engine. Candidate third structural claim (alongside no-win and adoption-routing); not yet ratified as never-patch canon.
Rams #10: world-update packs reuse ADR-0005's schedule format and the L9 data externalization; no new engine system — the league is a content cadence plus a frontier config value.
MaRo Surprise/Discovery: monthly rotation, not seed randomness, is the veteran freshness engine.

Interaction contract

Reads/writes: ADR-0005 seed schedules (packs are schedule entries; pdoom-data is the feedstock — EE-6 promoted, this is its customer), ADR-0009 fiction anchors (2017 start reaffirmed), ADR-0006 replay/versioning (runs are per-league artifacts; DQ-3 cross-version boards become league UX), L9 Balance separation (balance constants and world packs are distinct data surfaces on distinct cadences).

Rejected alternatives

Run-start rides the leagues (start ~present-day): loses the 2017 arc and the village-era early fiction. Rejected for the default mode; survives only as the parked present-month experiment.
Monolithic monthly world+balance patch: ambient-noise patches violate the patch-as-heartbeat philosophy. Decoupled instead.
Per-seed macro world regeneration (Civ-style map variance): fiction can't tolerate it (2017 start is semi-historical) and Pip doesn't need it — Factorio-mode accepted.

Amendments (2026-07-13, handover round)

League baseline = one seed (possibly one seed with a few variants — the elegant home for the parked present-month mode). One seed is promoted by default; players can generate and share their own — anyone can host a mini-league by posting a baseline seed. Community structure from seed-sharing, no new infrastructure.
Canon-ness ruled ~75% — bake deeply into config. Pip's reason: without the reality-tether the game drifts from its guiding star as "a tool / scenario builder / argument fosterer."

Consequences / open questions

Pipeline build (EE-6 promoted): structured monthly world-diff format an agent can draft and Pip approves. Estimate 1–2 days/month once built; the sustainability risk is real (solo liveops decay) — treat the pipeline as a product feature, not an aspiration.
League-notes format, seed-racer norms, what happens to in-flight runs at rollover (per-league artifacts suggest: they finish on their league's version).
Engagement dependency named by Pip: the cadence only pays if the game is public with real players "pretty darn soon" — sequencing pressure on L1..Ln.
The fixed 2017→present history grows by a month every league — the macro opening gets longer over time. Watch for opening fatigue; the parked present-month mode is the relief valve if it bites.