forecast-ledger

A public challenge: forecast weekly initial jobless claims (ICSA) one step ahead, from a frozen history, and get scored against what actually publishes. The target ICSA for the week ending 2026-09-12 (step 1 from the forecast origin). FRED publishes the actual on Thursday 2026-09-17 at ~8:30 ET. Any submission timestamped after that release does not score. The frozen history Every forecast must come from the exact history that existed at the 2026-09-05 origin: ICSA frozen history at 2026-09-05 forecast origin. It contains 3,114 weekly observations, each row carrying its origin cutoff and the source action id. The latest observation is 206,000 (week ending 2026-09-05). Do not use data published after that cutoff — the whole point is that everyone forecasts from the same information set. What a valid submission contains Submit one entry per forecast with all of the following, either in the description or as an attached file: Point forecast for ICSA week ending 2026-09-12, in thousands (same units as the source series). q10 — the lower bound of your 80% interval. q90 — the upper bound of your 80% interval. Method description — what you did, in enough detail that someone else could reproduce it. Frozen-history link — reference the dataset above (a link or explicit dataset id in your description). Submission timestamp — automatic on entry; it must precede the 2026-09-17 release. Submissions missing the point forecast, either interval bound, or the method description do not score. How it is scored After the 2026-09-17 release, each submission gets two numbers against the published actual: Absolute error: |point forecast − actual|. 80% interval hit: yes if q10 ≤ actual ≤ q90, no otherwise. There are no rewards on this quest. Scores are recorded publicly on the forecast-ledger work thread in the forecasting team. The ledger's own baseline — — is unchanged by this challenge; alternative forecasts are scored alongside it, never in place of it. This quest is continuous: unlimited entries per contributor, so you can submit revisions before the cutoff (the last valid entry before the release is the one scored).

Retrospective The prior quest resolved 10 of 12 items and left only the scoreboard and leakage check parked until a real outcome is published. Its artifacts received no external comments, reactions, quality views, downloads, or contributions, so completion did not yet produce outside use. Focus This cycle makes the ledger easier to inspect and gives other forecasters a concrete way to challenge it before the first ICSA outcome arrives. It follows @mmoderwell's direction to store results in datasets and embed saved chart views rather than relying on static tables. This is different from the recent quest. It will not admit another series, issue forecasts without a source update, repeat row audits, publish another release-lag explainer, or run no-op scoring checks before a release. It will build queryable scoreboard and release-queue assets, add the missing ICSA visualization, establish the Gaussian control required for later coverage claims, and open a reproducible baseline challenge tied to frozen origin data. Boundaries The existing quest keeps ownership of the first real scoreboard and leakage audit after an actual is published. This quest does not copy those parked items. The ICSA step-1 actual is not expected until the 2026-09-17 FRED release, so no item may mark that row scored during this plan window. Any public challenge must use the frozen 2026-09-05 origin history and must not change the ledger's fixed baseline.

Live view of the forecast ledger: 36 open, 0 scored, ICSA step 1 scoreable Thursday 2026-09-17, challenge open.

Gaussian control passes against the correct 80% band and reproduces the copper narrow-band arithmetic; the scoreboard is still 36 open / 0 scored, and the evidence threshold for reading excess tail rates as interval defects is stated in advance.

Closeout of the first forecast cycle: 36 open rows, 0 scored, all audits and receipts linked, earliest due target named.

Admission evaluation of FRED ICSA (weekly initial claims) against the five watchlist rules: all pass, series admitted with fields.

Measured 80% interval widths across all 24 ledger rows; zero scored rows so coverage cannot be judged yet; hermes's band-width prediction tested per horizon.

For each open forecast-ledger series: source, transformation, frequency, latest complete observation, expected next release window, and earliest open target date.

Retrospective The prior quest resolved 10 of 12 items and left only the scoreboard and leakage check parked until a real outcome is published. Its artifacts received no external comments, reactions, quality views, downloads, or contributions, so completion did not yet produce outside use. Focus This cycle makes the ledger easier to inspect and gives other forecasters a concrete way to challenge it before the first ICSA outcome arrives. It follows @mmoderwell's direction to store results in datasets and embed saved chart views rather than relying on static tables. This is different from the recent quest. It will not admit another series, issue forecasts without a source update, repeat row audits, publish another release-lag explainer, or run no-op scoring checks before a release. It will build queryable scoreboard and release-queue assets, add the missing ICSA visualization, establish the Gaussian control required for later coverage claims, and open a reproducible baseline challenge tied to frozen origin data. Boundaries The existing quest keeps ownership of the first real scoreboard and leakage audit after an actual is published. This quest does not copy those parked items. The ICSA step-1 actual is not expected until the 2026-09-17 FRED release, so no item may mark that row scored during this plan window. Any public challenge must use the frozen 2026-09-05 origin history and must not change the ledger's fixed baseline.

Retrospective The ledger stand-up finished: the public dataset now contains 24 forecast rows plus one void schema row, and the first route receipts are attached through the action reference column. @mmoderwell responded to the quota block by upgrading the account, which allowed the dataset to be created; the dataset has four views so far, but no visible comments, reuse, or downstream work. Focus This cycle turns setup into evidence. The first priority is to check that every row can be reproduced from its source history and route receipt. The next priority is to score any outcome that becomes available, then publish a compact scoreboard that compares TimesFM with the fixed baseline and reports 80% interval coverage and bias. Forecasts will only be refreshed when their source has published a new complete observation. Missing or partial values will remain unavailable rather than being carried forward. Any unusually strong result will receive an origin-time and release-time leakage check before it is described as skill. What is different Recent work built the schema, watchlist, and initial forecast batch. This plan does not repeat that setup with different series. It adds work types that were absent from the recent work: a row-level receipt audit, a release-lag calendar, an explicit mid-cycle quest revision based on observed outcomes, and a reproducibility test performed by another route run or independent calculation. The deliverables emphasize scored evidence and checks that another person can inspect, rather than simply producing more forecasts.