Methodology v0.1
method_version=0.1
This is the public scoring method for official forecasts on TrackRecord. It is not investment research and not a lawyer-reviewed document. We keep score. We do not play.
What TrackRecord scores
Wave 1 public scoreboard claims are official forecasts only: FOMC SEP, EIA STEO survey columns, and CPI / CES previous-print baselines versus the next official print. Analyst, pundit, and news claims are never scored. CME FedWatch and futures-implied paths are never scored. Regional nowcasts are not treated as “our” forecast.
Method version 0.1
- R1 — If a forecast reaches its horizon without hitting the target band, the result is
miss. - R2 — Earlier hits score higher (time decay). On a hit,
points = 1 + remaining/span, wherespanis calendar days from claim date to horizon andremainingis calendar days from the matching outcome as-of to the horizon. A miss scores0. Open claims havepoints = null. - R4 — Scoring is per-topic, not a career total across topics.
We do not use R3 (a consensus-deviation bonus). We do not publish Brier scores.
How a claim resolves
- Horizon (
horizon_at). Before that day the claim staysopenunless already within tolerance of a held matching outcome. - Matching outcome. When the horizon is reached, we look for a held official observation that matches the claim’s outcome kind (level, Q4 average, or Q4/Q4). If the horizon has passed and no matching outcome is held, the claim stays
open— it is not markedmiss. - Tolerance. A hit requires the absolute gap between target and outcome to be within the claim’s tolerance.
Outcome oracles by predictor
| Predictor | Outcome series | Notes |
|---|---|---|
| FOMC SEP (federal funds) | fed-effr | tolerance 0.125 |
| FOMC SEP (unemployment) | bls-cps-unrate | Q4 average when enough Q4 months are held |
| FOMC SEP (PCE) | — | Q4/Q4; monthly 12-month PCE is never used as the outcome → stays open |
| EIA STEO (retail gasoline) | eia-878-regular-gas | tolerance 0.05 |
| EIA STEO (crude stocks ex-SPR) | eia-wpsr-crude-ex-spr | tolerance 5000 thousand barrels |
| EIA STEO (gasoline stocks) | eia-wpsr-gasoline-stocks | tolerance 5000 thousand barrels |
| EIA STEO (crude production) | eia-wpsr-crude-prod | tolerance 250 thousand barrels/day |
| BLS prior-print (CPI) | bls-cpi-u | tolerance 0.05 |
| BLS prior-print (CES payrolls) | bls-ces-payrolls | tolerance 0 (naive no-change baseline, not a BLS staff forecast) |
| IMF WEO / OECD EO / ECB SPF / CBO / SCE / Focus / WASDE / OMB / OBR / AMECO / GACC | — | unscored_no_oracle until a lawful outcome series is minted |
Previous-print claims
A previous-print claim treats the adjacent earlier official print as a naive forecast of the next print. CPI uses tolerance 0.05 percentage points. CES payrolls use tolerance 0. Revisions follow the producer’s latest public vintage held in our observations.
Results
Each scored claim is one of: hit, miss, open, void, unscored_no_oracle.
A topic with no lawful outcome stays unscored. Unscored claims are not counted as misses on the public Correct / Missed / Open counters.
Hit-rate publication gates
We do not publish a hit-rate percentage when fewer than 20 expired scorable claims are held (n<20). Even when n≥20, a separate lawyer-clearance flag must be true before any percentage is rendered. Until both gates pass, pages may show Correct / Missed / Open counts only — never a hit-rate %.
Revisions and corrections
Producer revisions replace the Numbers we hold when the official vintage updates; the producer remains the authority. Presentation errors on TrackRecord (wrong as-of, broken source link, mis-rendered scoreboard row) follow the Corrections policy in the Terms of Service (Corrections section). We do not invent an alternate official print. Do not duplicate that procedure here.
What we do not score
- Analyst or pundit forecasts
- News articles or wire copy as outcomes or claims
- CME FedWatch / futures-implied policy paths
- Nowcasts marketed as TrackRecord’s own forecast
- Gold, Bitcoin, or WTI spot prices on the public Wave 1 scoreboard
Official producers remain the authority for the Numbers we score against. Verify on the topic page.