feat(forecast): fast-resolving commodity price bets — Gate-1 accelerator (#5233) by koala73 · Pull Request #5243 · koala73/worldmonitor

koala73 · 2026-07-12T10:08:20Z

Why

The energy pilot's EIA bets resolve weekly (deadline 2026-07-19), so bet_engine scorecard evidence was a week out. This adds a fast lane on the daily-updating commodities feed — bets resolve within the week (~4-day horizon → 2026-07-16) — so Gate-1 evidence lands sooner, and it proves the engine generalizes to a second feed with a different metric fn.

What

_bet-templates-commodities.mjs — bets over market:commodities-bootstrap:v1 for WTI (CL=F), Brent (BZ=F), nat-gas (NG=F), gold (GC=F). Directional (from the daily change), 2.5% threshold, 2-decimal prices, resolvable via the resolver's existing price(symbol==X) fn.
shapeResolutionFeed — the commodities feed is {_seed, data:{quotes:[...]}}, and the eval's iterateRecords only descends into array children, so the doubly-nested quotes were unreadable. Now exposed as the array. This also unblocks the pre-existing market commodity-price forecast path (it had the same read problem).
seed-forecast-bets.mjs — reads both feeds, emits both families. Commodity bets use the honest thin-history prior (0.4) until their own series accrues (Phase 2's ensemble replaces the probability regardless).

Evidence

8 new tests (templates, the shaper unwrap, and end-to-end YES/NO/pending through the real resolver). 870/870 forecast+bet suites green; api tsc exit 0.
Live prod preview — 8 bets total (4 energy @07-19, 4 commodity @07-16), all crisp + resolvable:
- Will the WTI crude oil price fall to at most 69.62 USD/bbl by 2026-07-16?
- Will the US natural gas price fall to at most 2.87 USD/MMBtu by 2026-07-16?
- Will the gold price fall to at most 4010.86 USD/oz by 2026-07-16?

Scope

Shadow lane only (invisible to users; held out of the skill Brier). Short-horizon price bets are ~coin-flip on direction by design — Gate 1 tests resolves + crisp + not-silly, not skill; calibration is Phase 2. A per-symbol commodity base-rate accumulator is a fast-follow (they're at the prior until then).

Refs #5233, #4930

https://claude.ai/code/session_01StNurp4TGC3JLHbTtKJhbp

@07-19

…or, #5233) The EIA energy bets resolve weekly (deadline 07-19). This adds a fast lane on the daily-updating commodities feed so bet_engine evidence lands within the week (~4-day horizon → 07-16) and proves the engine generalizes to a 2nd feed. - _bet-templates-commodities.mjs: crude (CL=F/BZ=F), nat-gas (NG=F), gold (GC=F) bets over market:commodities-bootstrap:v1. Directional (from the daily change), 2.5% threshold, 2-dp prices, resolvable via the existing price(symbol==X) fn. - seed-forecast-resolutions.mjs: shapeResolutionFeed now exposes the nested {data:{quotes:[...]}} array — the eval's iterateRecords only descends into array children, so the doubly-nested quotes were unreadable. This ALSO unblocks the pre-existing market commodity-price forecast resolution path. - seed-forecast-bets.mjs: reads both feeds, emits both families. Commodity bets use the honest thin-history prior (0.4) until their own series accrues. Prod preview: 8 bets total (4 energy @07-19, 4 commodity @07-16), all crisp + resolvable, e.g. "Will the WTI crude oil price fall to at most 69.62 USD/bbl by 2026-07-16?". 8 new tests (templates, shaper, end-to-end YES/NO/pending); 870/870 forecast+bet green; api tsc exit 0. Refs #5233, #4930 Claude-Session: https://claude.ai/code/session_01StNurp4TGC3JLHbTtKJhbp

strix-security · 2026-07-12T10:08:25Z

Strix Security Review

No security issues found.

_{Updated for 518add3.}

Reviewed by Strix
_{Re-run review · Configure security review settings}

vercel · 2026-07-12T10:08:26Z

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

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
worldmonitor	Ignored	Preview	Jul 12, 2026 1:03pm

greptile-apps · 2026-07-12T10:14:08Z

Greptile Summary

This PR adds a fast-resolving commodity price bet lane to the shadow bet_engine, complementing the existing weekly EIA energy bets with daily-updating WTI, Brent, nat-gas, and gold bets that resolve within 4 days. It also fixes a feed-shape bug in shapeResolutionFeed that was silently blocking resolution for the pre-existing commodity-price forecast path.

New _bet-templates-commodities.mjs: four commodity symbols, directional bets derived from the feed's daily change field, 2.5% move threshold, 2-decimal rounding, metricKey compatible with the resolver's existing price() function.
shapeResolutionFeed fix: unwraps the doubly-nested {_seed, data:{quotes:[...]}} envelope to the bare quotes array so iterateRecords can reach individual quote objects; without this the price(symbol==X) path would always VOID.
seed-forecast-bets.mjs: merges both template families and both feed keys; commodity bets receive the honest thin-history prior (0.4) until a per-symbol accumulator is built in Phase 2.

Confidence Score: 4/5

Shadow-lane only change with no user-facing writes; the core fix is correct and well-tested.

The shapeResolutionFeed fix and template wiring are solid. Two non-blocking issues: stale energy labels in the log path and the silent default of absent/zero change to rising direction, which could skew the Gate-1 evidence pool on days the commodity feed omits the field.

scripts/_bet-templates-commodities.mjs (zero/absent change handling) and the log messages in scripts/seed-forecast-bets.mjs.

Important Files Changed

Filename	Overview
scripts/_bet-templates-commodities.mjs	New commodity bet template file defining 4 symbols with directional price bets over a 4-day horizon; direction derived from daily change field, defaulting to rising when absent or zero.
scripts/seed-forecast-bets.mjs	Adds commodity bet family to the seeder; merges both template families and feed keys; stale energy labels remain in two log lines.
scripts/seed-forecast-resolutions.mjs	Adds shapeResolutionFeed branch for the commodities feed, unwrapping the double-nested envelope into the bare quotes array so price(symbol==X) resolution works correctly.
tests/bet-templates-commodities.test.mjs	8 new tests covering template generation, shaper unwrapping, end-to-end YES/NO/pending resolution, and seeder integration; fixture and assertions are accurate.

Sequence Diagram

%%{init: {'theme': 'neutral'}}%%
sequenceDiagram
    participant Cron as Railway Cron
    participant Seeder as seed-forecast-bets.mjs
    participant Redis as Redis (Upstash)
    participant EIATemplate as ENERGY_BET_TEMPLATES
    participant CommodityTemplate as COMMODITY_BET_TEMPLATES
    participant BaseRate as baseRateProbability
    participant Resolver as seed-forecast-resolutions.mjs

    Cron->>Seeder: trigger daily run
    Seeder->>Redis: GET energy:eia-petroleum:v1
    Seeder->>Redis: GET market:commodities-bootstrap:v1
    Seeder->>Redis: GET forecast:bets:eia-series:v1
    Redis-->>Seeder: raw feeds + prior series
    Seeder->>Seeder: unwrapFeeds() strips envelope
    Seeder->>EIATemplate: extractMetric(unwrapped EIA feed)
    EIATemplate-->>Seeder: 4 energy bets
    Seeder->>CommodityTemplate: extractMetric(unwrapped commodities feed)
    CommodityTemplate-->>Seeder: 4 commodity bets
    Seeder->>BaseRate: baseRateProbability(EIA series, resolution)
    BaseRate-->>Seeder: accumulator-backed probability
    Seeder->>BaseRate: baseRateProbability([], resolution)
    BaseRate-->>Seeder: 0.4 thin-history prior
    Seeder->>Redis: LPUSH forecast:bets:history:v1
    Seeder->>Redis: SET forecast:bets:eia-series:v1
    Note over Resolver: Resolution run at/after deadline
    Resolver->>Redis: GET market:commodities-bootstrap:v1
    Resolver->>Resolver: shapeResolutionFeed() unwraps to quotes[]
    Resolver->>Resolver: "price(symbol==CL=F) crossesThreshold?"
    Resolver-->>Redis: YES/NO/VOID

%%{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"}}}%%
sequenceDiagram
    participant Cron as Railway Cron
    participant Seeder as seed-forecast-bets.mjs
    participant Redis as Redis (Upstash)
    participant EIATemplate as ENERGY_BET_TEMPLATES
    participant CommodityTemplate as COMMODITY_BET_TEMPLATES
    participant BaseRate as baseRateProbability
    participant Resolver as seed-forecast-resolutions.mjs

    Cron->>Seeder: trigger daily run
    Seeder->>Redis: GET energy:eia-petroleum:v1
    Seeder->>Redis: GET market:commodities-bootstrap:v1
    Seeder->>Redis: GET forecast:bets:eia-series:v1
    Redis-->>Seeder: raw feeds + prior series
    Seeder->>Seeder: unwrapFeeds() strips envelope
    Seeder->>EIATemplate: extractMetric(unwrapped EIA feed)
    EIATemplate-->>Seeder: 4 energy bets
    Seeder->>CommodityTemplate: extractMetric(unwrapped commodities feed)
    CommodityTemplate-->>Seeder: 4 commodity bets
    Seeder->>BaseRate: baseRateProbability(EIA series, resolution)
    BaseRate-->>Seeder: accumulator-backed probability
    Seeder->>BaseRate: baseRateProbability([], resolution)
    BaseRate-->>Seeder: 0.4 thin-history prior
    Seeder->>Redis: LPUSH forecast:bets:history:v1
    Seeder->>Redis: SET forecast:bets:eia-series:v1
    Note over Resolver: Resolution run at/after deadline
    Resolver->>Redis: GET market:commodities-bootstrap:v1
    Resolver->>Resolver: shapeResolutionFeed() unwraps to quotes[]
    Resolver->>Resolver: "price(symbol==CL=F) crossesThreshold?"
    Resolver-->>Redis: YES/NO/VOID

Comments Outside Diff (1)

scripts/seed-forecast-bets.mjs, line 142-148 (link)

Stale "energy" labels in log output

The two log lines still refer to "energy bet(s)" after adding the commodity family. If the EIA feed is absent (so only commodity bets are produced), ops will see "published N shadow energy bet(s)" for what are purely commodity forecasts — and the warning path says "no energy bets generated" even when the real reason is that both feeds were empty. Inaccurate labels make on-call triage harder.

_{Reviews (1): Last reviewed commit: "feat(forecast): fast-resolving commodity..." | Re-trigger Greptile}

greptile-apps · 2026-07-12T10:14:13Z

+        kind: 'hard',
+        metricKey: `${COMMODITY_FEED}|price(symbol==${symbol})`,
+        operator: 'crosses',


Zero change always generates a "rising" bet

wantUp = changePct >= 0, so when change is absent from the feed (mapped to 0 by extractMetric) or exactly 0.0 (a flat day), the template always mints a "rise to at least" bet. A missing or zero daily-change field is a plausible feed state — Yahoo Finance futures data occasionally omits change on non-trading sessions — and the resulting bias silently skews the Gate-1 evidence pool toward bullish direction for that day's run. Consider using a null sentinel to skip bet generation when the daily direction is undetermined rather than defaulting to up.

…d (PR #5243 review) Greptile: wantUp = changePct >= 0 defaulted an absent or exactly-flat `change` to a "rise" bet, biasing the shadow pool bullish on non-trading sessions (Yahoo omits `change` then). Now extractMetric keeps a null sentinel for absent change, and buildResolutionSpec emits no bet when direction is undetermined (null or 0) — no silent bullish default. Test: absent/flat change → 0 bets. 9 commodity tests, 871/871 forecast+bet green; api tsc exit 0. Refs #5233, #5243 Claude-Session: https://claude.ai/code/session_01StNurp4TGC3JLHbTtKJhbp

…eeder logs (PR #5243 review) P2 (correctness) — stale kept-warm quote could resolve as a deadline price. The commodities feed uses extendExistingTtl on fetch failure, preserving the OLD _seed.fetchedAt, but shapeResolutionFeed discarded it — so a quote fetched days before the deadline scored a false YES/NO (reproduced). Now shapeResolutionFeed stamps each quote with the envelope fetchedAt as `asOf`, and the eval's settlement gate (previously value-only) also covers the `price` fn: a quote dated before the deadline day pends, and VOIDs after the 10d grace. This also hardens the pre-existing market commodity-price forecast path. P3 (ops) — the combined seeder logged "shadow energy bet(s)" / "no energy bets" even on commodity-only runs. Now family-neutral with a per-domain breakdown (e.g. "published 8 shadow bet(s) [energy:4, market:4]"). Tests: stale-envelope pending + VOID-after-grace regressions, asOf stamping. 873/873 forecast+bet green; api tsc exit 0. Refs #5233, #5243 Claude-Session: https://claude.ai/code/session_01StNurp4TGC3JLHbTtKJhbp

…5243 review) P1 (correctness) — a stale post-deadline sample survived feed recovery and was scored over the fresh quote. samplePendingEntries stamped samples with the CYCLE time, so a stale kept-warm reading got a post-deadline ts; the next fresh cycle then (a) skipped sampling (a post-deadline sample already existed) and (b) had the at-deadline path prefer that stored stale value. Now samples are stamped with the SOURCE observation time (asOf) via a new extractMetricObservation, so a stale reading carries a pre-deadline ts, is never selected, and never blocks the fresh quote. Two-cycle stale→fresh regression added. P2 #2 — a partial refresh that dropped a symbol VOIDed the bet immediately (missing record → gate fell through → no_establishable_metric). Now a present feed with an absent matched record pends through the settlement grace and VOIDs only if it never returns. P2 #3 — the seeder could mint a newly dated bet from a stale kept-warm envelope. A per-feed generation freshness contract (commodities: 5d, tolerates any weekend/holiday) drops a feed whose _seed.fetchedAt is beyond it, so no bet is generated from stale prices. Also: the settlement gate is now scoped to POINT windows (at-deadline) — a within-horizon spec resolves from the asOf-stamped sample timeline, not the current feed record, and must not be gated on it (caught by an existing test). 876/876 forecast+bet green; api tsc exit 0. Refs #5233, #5243 Claude-Session: https://claude.ai/code/session_01StNurp4TGC3JLHbTtKJhbp

…regression (#5327) * fix(forecast): present() observation must be 0 (not NaN) on absence — VOID-rate regression While triaging the 68% resolution VOID rate, found a regression I introduced in the sampling refactor (#5243 P1): extractMetricObservation returned NaN for a missing matched record, but present() is a boolean existence check — an absent subject means the event did NOT occur → 0, not NaN. samplePendingEntries then stored an error sample instead of a valid 0, so a legitimately-NO within-horizon present spec (e.g. infra:outages present(country==X) for a quiet country) stranded with an empty timeline and VOIDed as no_establishable_metric instead of resolving NO. Now mirrors extractMetricValue's `record ? 1 : 0`. Tests: extractMetricObservation present-absent → 0 (matches extractMetricValue); present-within-horizon resolves NO on a sampled absence. 908/908 forecast+bet green; api tsc exit 0. Refs #5233, #4930 Claude-Session: https://claude.ai/code/session_01StNurp4TGC3JLHbTtKJhbp * fix(review): assert present observation timestamp

vercel Bot deployed to Preview July 12, 2026 10:11 View deployment

greptile-apps Bot reviewed Jul 12, 2026

View reviewed changes

koala73 added 3 commits July 12, 2026 14:20

koala73 merged commit 16c31bc into main Jul 12, 2026
26 checks passed

koala73 mentioned this pull request Jul 15, 2026

fix(forecast): present() observes 0 (not NaN) on absence — VOID-rate regression #5327

Merged

koala73 mentioned this pull request Jul 23, 2026

feat(forecast): Phase 2 re-engine — LLM ensemble + calibration on prediction-markets + FRED (#4930 Bet 4) #5525

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(forecast): fast-resolving commodity price bets — Gate-1 accelerator (#5233)#5243

feat(forecast): fast-resolving commodity price bets — Gate-1 accelerator (#5233)#5243
koala73 merged 4 commits into
mainfrom
feat/forecast-reengine-fast-commodity-bets

koala73 commented Jul 12, 2026

Uh oh!

strix-security Bot commented Jul 12, 2026 •

edited

Loading

Uh oh!

vercel Bot commented Jul 12, 2026 •

edited

Loading

Uh oh!

greptile-apps Bot commented Jul 12, 2026 •

edited

Loading

Comments Outside Diff (1)

Uh oh!

greptile-apps Bot Jul 12, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

koala73 commented Jul 12, 2026

Why

What

Evidence

Scope

Uh oh!

strix-security Bot commented Jul 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Strix Security Review

Uh oh!

vercel Bot commented Jul 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

greptile-apps Bot commented Jul 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Greptile Summary

Confidence Score: 4/5

Important Files Changed

Sequence Diagram

Comments Outside Diff (1)

Uh oh!

greptile-apps Bot Jul 12, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

strix-security Bot commented Jul 12, 2026 •

edited

Loading

vercel Bot commented Jul 12, 2026 •

edited

Loading

greptile-apps Bot commented Jul 12, 2026 •

edited

Loading