Skip to content
Canli Capital

Research

Pre-FOMC announcement drift — no-return feasibility protocol

By , Canli Capital

Declared: 2026-08-16
Return data: unopened
Hypotheses spent: zero

Question

Can the scheduled FOMC decision calendar and exact statement-release clock be reconstructed from official Federal Reserve sources without using market prices or learning which events were profitable?

Locked audit

  • Official Federal Reserve current calendar for 2021 onward.
  • Official per-year historical meeting pages for 2016–2020.
  • Regular meetings only; emergency conference calls and unscheduled actions are excluded.
  • Exact HTML statement URL, decision date, release-time text, source hash, and retrieval lineage.
  • No intraday bars, closes, spreads, returns, signals, positions, or performance statistics.

Gates before any return identity

  1. Eight regular scheduled decisions in ordinary complete years. The March 17–18, 2020 regular meeting was superseded during the emergency period; the unscheduled March 2/3 and March 15 actions are excluded because a pre-announcement window was not knowable from the regular calendar. The resulting statement-event corpus should contain 79 decisions: eight in nine years and seven in 2020.
  2. Statement release-time metadata coverage must be 100%.
  3. The 2:00 p.m. ET release convention must be measured, not assumed.
  4. Point-in-time schedule revisions/confirmations must be reconstructable. The current Federal Reserve calendar says each future meeting date is tentative until confirmed at the immediately preceding meeting; an ex-post list of final dates is not sufficient lineage.

Failure of gate 4 yields CALENDAR_LINEAGE_REQUIRED, not permission to open returns.

Resolved schedule lineage

The follow-up audit binds ten official annual schedule press releases for 2016–2025. Each page contains eight target-year meeting slots, is SHA-256 bound, and was published before the target year. The minimum lead is 117 XNYS sessions. All 79 completed regular decisions match those 80 initially scheduled slots.

The sole schedule-only slot is March 17–18, 2020. The official 2020 history marks it cancelled, and the March 15 emergency statement is bound at 5:00 p.m. EDT, two XNYS sessions before the scheduled decision date. Dropping this event would create ex-post selection. Any return identity must preserve it as CANCELLED_NO_ENTRY_OR_FLATTEN_AT_NEXT_ELIGIBLE_OPEN.

Decision: PASS_TO_RETURN_PREREGISTRATION. This is not return evidence. The single identity is locked separately in PREREG_PRE_FOMC_ANNOUNCEMENT_DRIFT.md; returns remain unopened until the market-input manifest and runner enforce that contract.

Market-data readiness result

The locked event/control expansion produces 312 valid matched-control windows and 782 unique XNYS session files. Every required object is visible in the Polygon SIP quote archive; no required date is missing, and every completed event retains at least three controls. The cancelled March 2020 slot has no market window.

The available provider routes nevertheless fail implementation feasibility:

  • The required full-day SIP quote files total 3,005.294 GiB compressed, far beyond the locked 50 GiB ceiling for a bounded single-instrument research ingest.
  • The configured flat-file credentials can list objects but return HTTP 403 to one-byte endpoint retrieval probes.
  • The server-filtered Polygon historical SPY quote endpoint returns HTTP 403 NOT_AUTHORIZED.
  • Databento consolidated US-equity depth begins in 2023 and cannot cover the frozen 2016–2025 panel; single-venue history beginning in 2018 is not a substitute for consolidated NBBO.

Decision: DATA_GATED. No quote record was decompressed or parsed, no return was computed, and no hypothesis was spent. Progress requires a server-filtered historical SIP quote entitlement (or a vendor-supplied SPY-only export) covering 2015-12-29 through 2025-12-10 with bid, ask, sizes, timestamps, conditions, sequence/order lineage, and corrections. Alpaca shadow data cannot replace this historical research feed.