fix(forecast): resolution-spec emission amendment — horizon-commensurable count + resolvable supply/gps windows (#5010) by koala73 · Pull Request #5012 · koala73/worldmonitor

koala73 · 2026-07-07T08:19:37Z

Closes #5010 (Bet-1 follow-up; unblocks #5007). Epic #4930.

Why now: the Bet-2 planning review found the frozen spec shapes left only ~1 currently-published forecast resolvable. Amending emission one day into the resolution-spec clock resets a near-empty record; every day of delay makes the reset more expensive.

Changes (emission only — `scripts/_forecast-resolution.mjs`)

conflict/ucdp count thresholds are horizon-commensurable. The signal tally is a 365-day trailing count (TRAILING_WINDOW_MS, seed-ucdp-events.mjs), so a raw threshold systematically resolves NO over a ~30d horizon — a biased Brier. Now: threshold = max(1, round(tally × horizonMs/365d × CONFLICT_ESCALATION_RATIO)) (named, exported knob, like MARKET_PRICE_MOVE_RATIO), counted over new events dated within [emission, deadline] — the count() semantics are documented in the grammar header.
supply_chain/gps windows: at-deadline point reads instead of a sustained-48h condition that is unestablishable from current-snapshot feeds (forced permanent VOID).
Left as shipped (confirmed fine): prediction_market: at-endDate, market/infrastructure: within-horizon.

Guardrails carried: omission semantics preserved (no new fields, no nulls), grammar unchanged and still parseable (spaces/?/) values), signal-vocab + horizon drift guards intact.

Verification

Tests 54/54 (+3 amendment tests: at-deadline windows for both families; a source-scan asserting no sustained window token is emitted; horizon-scaling of the same tally across 24h vs 30d). Benchmark 6/6; history 15/15 and trace-export 310/310 untouched-green; biome clean.
Live DoD spot-check (fresh detector batch, 49 hard specs): 100% shape-resolvable (target ~90%), window distribution within-horizon: 41, at-deadline: 7, at-endDate: 1, zero sustained; conflict thresholds horizon-scoped on real data (7d: 15/1/5 instead of raw 365d tallies).
D2 no-selection-change gate re-passed: real selection path with/without the enrichment pass → identical published id set and order.

Per the issue: this intentionally supersedes the ~1 day of old-shape specs in history — the clock resets to today with resolvable shapes.

…able count + resolvable supply/gps windows (#5010) Bet-2 planning found the frozen Bet-1 spec shapes left only ~1 currently-published forecast resolvable. Amending emission now, one day into the clock, resets a near-empty record instead of a month-old one. 1. conflict/ucdp count thresholds are horizon-commensurable: the signal tally is a 365-day trailing count, so a raw threshold would systematically resolve NO over a ~30d window (biased Brier). The threshold is now max(1, round(tally × horizonMs/365d × CONFLICT_ESCALATION_RATIO)) — a named, exported knob like MARKET_PRICE_MOVE_RATIO — counted over NEW events dated within [emission, deadline]. count() semantics documented in the grammar header. 2. supply_chain/gps windows change from a sustained-48h condition (which is unestablishable from current-snapshot feeds and forced permanent VOID) to an at-deadline point read the Bet-2 resolver can establish. prediction_market at-endDate and market/infrastructure within-horizon stay as shipped (confirmed fine). Omission semantics preserved — no new fields, no nulls. Emission only: no probability-spine or publish-selection change (D2 gate re-passed: identical published set and order). Live spot-check (fresh detector batch, 49 hard specs): 100% shape- resolvable, zero sustained windows, conflict thresholds horizon-scoped (7d: 15/1/5 vs raw tallies). Tests 54/54; benchmark 6/6. Closes #5010 🤖 Generated with Claude Code Co-Authored-By: Claude <[email protected]>

vercel · 2026-07-07T08:19:43Z

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
worldmonitor	Ignored	Preview	Jul 7, 2026 8:26am

greptile-apps · 2026-07-07T08:23:48Z

Greptile Summary

This PR fixes two classes of permanently-unresolvable forecast spec shapes introduced in the Bet-1 emission: it replaces the sustained-48h window on supply_chain/gps with at-deadline point reads (snapshot feeds cannot establish a sustained condition), and it scales the conflict/ucdp_zone event-count threshold from the raw 365-day trailing tally to the forecast horizon, eliminating the systematic NO bias over short windows.

CONFLICT_ESCALATION_RATIO and YEAR_MS are added; the threshold now scales as max(1, round(tally × horizonMs/365d × ratio)), matching the semantics of count() which counts new events in [emission, deadline].
FAMILY_WINDOW entries for supply_chain and gps change from 'sustained-48h' to 'at-deadline'; all other families are untouched.
Three new tests cover the window changes, the absence of any sustained-* token, and the horizon-proportional threshold scaling across 24h vs 30d.

Confidence Score: 4/5

Safe to merge; changes are emission-only, deterministic, and covered by 54 tests including three new amendment-specific assertions.

Both changed families are well-tested with explicit before/after assertions and a source-scan guard. The only open item is that YEAR_MS is private, leaving test literals to duplicate the constant rather than reference the single source of truth.

scripts/_forecast-resolution.mjs — specifically the YEAR_MS constant and the horizon-scaling formula in deriveHardMetrics.

Important Files Changed

Filename	Overview
scripts/_forecast-resolution.mjs	Core emission module: adds CONFLICT_ESCALATION_RATIO + YEAR_MS, switches supply_chain/gps window to at-deadline, and scales conflict/ucdp_zone count threshold to horizon. Logic is correct; YEAR_MS is module-private while tests duplicate the literal.
tests/forecast-resolution.test.mjs	New describe block covers at-deadline window assertions, source-scan guard for sustained-* tokens, and dual-horizon threshold scaling; updated existing threshold assertion to match the new formula.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["deriveHardMetrics - conflict/ucdp_zone"] --> B{"firstFiniteSignalCount ucdp or conflict_events"}
    B -- "null" --> C["return null - judged spec"]
    B -- "tally" --> D{"HORIZON_MS lookup"}
    D -- "undefined" --> E["return null - judged spec"]
    D -- "horizonMs" --> F["threshold = max(1, round(tally x horizonMs/YEAR_MS x ESCALATION_RATIO))"]
    F --> G["Hard spec - window: within-horizon"]
    H["deriveHardMetrics - supply_chain"] --> I["threshold = 60"]
    I --> J["Hard spec - window: at-deadline"]
    K["deriveHardMetrics - gps"] --> L{"firstFiniteSignalCount gps_jamming"}
    L -- "null" --> M["return null - judged spec"]
    L -- "hexes" --> N["threshold = max(1, round(hexes))"]
    N --> O["Hard spec - window: at-deadline"]

%%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%%
flowchart TD
    A["deriveHardMetrics - conflict/ucdp_zone"] --> B{"firstFiniteSignalCount ucdp or conflict_events"}
    B -- "null" --> C["return null - judged spec"]
    B -- "tally" --> D{"HORIZON_MS lookup"}
    D -- "undefined" --> E["return null - judged spec"]
    D -- "horizonMs" --> F["threshold = max(1, round(tally x horizonMs/YEAR_MS x ESCALATION_RATIO))"]
    F --> G["Hard spec - window: within-horizon"]
    H["deriveHardMetrics - supply_chain"] --> I["threshold = 60"]
    I --> J["Hard spec - window: at-deadline"]
    K["deriveHardMetrics - gps"] --> L{"firstFiniteSignalCount gps_jamming"}
    L -- "null" --> M["return null - judged spec"]
    L -- "hexes" --> N["threshold = max(1, round(hexes))"]
    N --> O["Hard spec - window: at-deadline"]

_{Reviews (1): Last reviewed commit: "fix(forecast): resolution-spec emission ..." | Re-trigger Greptile}

greptile-apps · 2026-07-07T08:23:51Z

+export const CONFLICT_ESCALATION_RATIO = 1.5;
+const YEAR_MS = 365 * 24 * 60 * 60 * 1000;


YEAR_MS is module-private, so the two test assertions that verify the threshold formula must inline 365 * 24 * 60 * 60 * 1000 literally. If the constant is ever changed, tests would diverge silently. Exporting it would let tests reference the same value the implementation uses, following the pattern established by CONFLICT_ESCALATION_RATIO and HORIZON_MS.

Suggested change

export const CONFLICT_ESCALATION_RATIO = 1.5;

const YEAR_MS = 365 * 24 * 60 * 60 * 1000;

export const CONFLICT_ESCALATION_RATIO = 1.5;

export const YEAR_MS = 365 * 24 * 60 * 60 * 1000;

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

…#5010 (cherry-picked from #5011)

vercel Bot deployed to Preview July 7, 2026 08:23 View deployment

greptile-apps Bot reviewed Jul 7, 2026

View reviewed changes

test(forecast): assert within-horizon/at-endDate families unchanged by …

0abfb1e

…#5010 (cherry-picked from #5011)

koala73 enabled auto-merge (squash) July 7, 2026 08:26

koala73 mentioned this pull request Jul 7, 2026

fix(forecast): horizon-commensurable count + point-read supply/gps windows (#5010) #5011

Closed

koala73 merged commit 5f0427c into main Jul 7, 2026
23 checks passed

koala73 mentioned this pull request Jul 9, 2026

epic(forecast): verified probabilistic forecasting — resolution engine, public Brier track record, learned calibration, LLM ensemble (MiroFish++ roadmap) #4930

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(forecast): resolution-spec emission amendment — horizon-commensurable count + resolvable supply/gps windows (#5010)#5012

fix(forecast): resolution-spec emission amendment — horizon-commensurable count + resolvable supply/gps windows (#5010)#5012
koala73 merged 2 commits into
mainfrom
fix/5010-resolution-spec-emission

koala73 commented Jul 7, 2026

Uh oh!

vercel Bot commented Jul 7, 2026 •

edited

Loading

Uh oh!

greptile-apps Bot commented Jul 7, 2026

Uh oh!

greptile-apps Bot Jul 7, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

		export const CONFLICT_ESCALATION_RATIO = 1.5;
		const YEAR_MS = 365 * 24 * 60 * 60 * 1000;

Conversation

koala73 commented Jul 7, 2026

Changes (emission only — scripts/_forecast-resolution.mjs)

Verification

Uh oh!

vercel Bot commented Jul 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

greptile-apps Bot commented Jul 7, 2026

Greptile Summary

Confidence Score: 4/5

Important Files Changed

Flowchart

Uh oh!

greptile-apps Bot Jul 7, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Changes (emission only — `scripts/_forecast-resolution.mjs`)

vercel Bot commented Jul 7, 2026 •

edited

Loading