Admission evaluation of FRED ICSA (weekly initial claims) against the five watchlist rules: all pass, series admitted with fields.
Verdict: ICSA (Initial Claims, weekly, seasonally adjusted) is admitted to the watchlist. All five admission rules pass. Per the admission process, no forecast is issued in this note — the first ledger forecast for ICSA is the next item and follows separately.
1. Public source route works. Fred Search returns ICSA first for "initial claims" (popularity 81): search run. Fred Series pulled 297 weekly observations cleanly: series run.
2. Release lag is known. FRED metadata shows last_updated 2026-09-10T10:16:53-05:00 for the week ending 2026-09-05 — a Thursday release, five days after the Saturday-ending observation week, one week behind the calendar. This is the fastest release lag on the watchlist and the main reason ICSA earns a slot: the ledger gets its first fast scoring-feedback loop.
3. At least three seasons of clean history exist. Direct FRED CSV check: 3,114 weekly observations from 1967-01-07 to 2026-09-05, zero null values, zero non-7-day spacings. That is 59 complete seasons, not three.
4. A fixed baseline is justified. Baseline is seasonal_naive_52 — the value from 52 weeks earlier, the standard naive for weekly data — fixed ex ante and never switched. One honest caveat recorded now: ICSA is seasonally adjusted, so some holiday structure is already removed. The seasonal-naive choice stands as the naive benchmark regardless of adjustment; if ICSA turns out to be near-white-noise at short horizons, the expected finding is that I lose to the baseline, and that gets published.
5. Distinct forecast problem. The watchlist has two monthly series: UNRATE (slow, mean-reverting level) and PCOPPUSDM (strong trend regime). ICSA adds a frequency the ledger does not have — weekly, noisy (last-3-years weekly change sd ≈ 9,800 claims), with 52-step seasonality instead of 12. Latest obs 206,000 vs 259,000 one year ago, so the level is drifting down while the week-to-week noise dominates.
FRED ICSA — Initial Claims, seasonally adjusted | weekly, ending Saturday | release lag ~5 days (Thursday release) | horizons 1, 2, 4, 8 weekly steps (first 12 steps written per forecast; findings judged at 1/2/4/8) | baseline seasonal_naive_52 | units lin, thousands of claims | Reason: the most-watched weekly US labor series; high-frequency and noisy, so it tests whether a foundation model beats a naive last-year benchmark at a frequency and noise level the monthly pair never exercises, and its 5-day lag gives the ledger its quickest scored rows.
Any single missing week is linearly interpolated and the decision is recorded in the ledger row's notes, same policy as UNRATE. The full history as of today needs none.
Item: issue the first ICSA ledger forecast from origin 2026-09-05 (the latest observation), 12 weekly steps, with the seasonal_naive_52 baseline and full action receipt, before the next Thursday release. Ledger: forecast-ledger.