OPEN RESEARCH / A RECORD THAT KEEPS SCORE
Every forecast leaves a record.
See what was predicted, which information was available and what happened afterward. Original predictions stay unchanged.
RECENT ARCHIVE / EVERY ELIGIBLE OUTCOME
The scorecard, including the misses.
Results stay separated by series, units, geography and forecast horizon. Every valid observed prediction in the returned archive contributes to its group, including ties and misses.
No scored outcomes yet. A pending forecast is not an accuracy result.
36 awaiting outcome · 4 unavailable · 0 malformed forecast records excluded.
How this scorecard is calculated
Absolute errors are recomputed from the preserved prediction, last value at issue and returned target observation. MAE is the arithmetic mean of those absolute errors. Wins have lower error than no change; differences within 0.0000000001 source units count as ties. The edition summary and row labels use this same calculation. Separate issue dates can predict the same target month, so these counts are forecasts rather than independent target months. Scores use the latest source revisions returned by the archive. They cover the returned editions from the 100 most recent stored records, not an assertion that all historical records fit on this page. This does not establish purchasing savings or a probability interval.
FORECAST EDITION
The issued prediction
Issued: Oct 4, 2026, 5:58 PM
Source retrieved at issue: Oct 4, 2026, 4:56 PM
Scoring source retrieved: Oct 11, 2026, 8:43 PM
| Series / units | Observed → target | Forecast / no change | Actual / absolute error | Outcome status |
|---|---|---|---|---|
| Softwood lumberIndex 1982=100 · 3 months | Aug 2026 → Nov 2026 | 305.18 / 294.9Damped 12-month log trend | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| Softwood lumberIndex 1982=100 · 6 months | Aug 2026 → Feb 2027 | 294.9 / 294.9Last observed value | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| Ready-mix concreteIndex 1982=100 · 3 months | Aug 2026 → Nov 2026 | 395.19 / 395.19Last observed value | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| Ready-mix concreteIndex 1982=100 · 6 months | Aug 2026 → Feb 2027 | 398.47 / 395.19Damped 12-month log trend | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| Hydraulic cementIndex 1982=100 · 3 months | Aug 2026 → Nov 2026 | 355.67 / 355.67Last observed value | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| Hydraulic cementIndex 1982=100 · 6 months | Aug 2026 → Feb 2027 | 355.67 / 355.67Last observed value | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| Gypsum productsIndex 1982=100 · 3 months | Aug 2026 → Nov 2026 | 484.84 / 484.84Last observed value | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| Gypsum productsIndex 1982=100 · 6 months | Aug 2026 → Feb 2027 | 484.58 / 484.84Damped 12-month log trend | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| Steel mill productsIndex 1982=100 · 3 months | Aug 2026 → Nov 2026 | 399.15 / 381.16Damped 12-month log trend | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| Steel mill productsIndex 1982=100 · 6 months | Aug 2026 → Feb 2027 | 408.69 / 381.16Damped 12-month log trend | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| Copper wire and cableIndex Dec 1986=100 · 3 months | — → — | — / —Monthly observations must be contiguous, unique and in ascending order. Missing months are not filled. | — | Forecast unavailable |
| Copper wire and cableIndex Dec 1986=100 · 6 months | — → — | — / —Monthly observations must be contiguous, unique and in ascending order. Missing months are not filled. | — | Forecast unavailable |
| Asphalt and tar roofing/sidingIndex 1982=100 · 3 months | Aug 2026 → Nov 2026 | 357.92 / 374.1Same month last year | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| Asphalt and tar roofing/sidingIndex 1982=100 · 6 months | Aug 2026 → Feb 2027 | 378.19 / 374.1Damped 12-month log trend | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| Mineral-wool envelope insulationIndex 1982=100 · 3 months | Aug 2026 → Nov 2026 | 272.37 / 272.37Last observed value | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| Mineral-wool envelope insulationIndex 1982=100 · 6 months | Aug 2026 → Feb 2027 | 272.37 / 272.37Last observed value | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| U.S. new single-family home salesThousands of homes, annual rate · 3 months | Aug 2026 → Nov 2026 | 684 / 684Last observed value | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| U.S. new single-family home salesThousands of homes, annual rate · 6 months | Aug 2026 → Feb 2027 | 630 / 684Same month last year | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| U.S. new-home months of supplyMonths of supply · 3 months | Aug 2026 → Nov 2026 | 8.5 / 8.5Last observed value | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
| U.S. new-home months of supplyMonths of supply · 6 months | Aug 2026 → Feb 2027 | 8.5 / 8.5Last observed value | — | Awaiting source outcomeAwaiting an exact target-month observation with matching units and geography. |
The record starts with a real issue date.
These forecasts were published before their target months. Their eventual accuracy will appear when the target observations become available. Historical backtests are separate.
Inspect the historical model testWhat is preserved—and how scoring works
Each edition stores the full retrieved source history, the selected model, the original prediction and its historical error range. The earliest durable record in a UTC calendar month is used for that month's edition. If simultaneous requests create duplicates, the archive selects by storage creation time; every original record remains unchanged. Opening the archive after a fresh source retrieval can issue that month's edition; this is not a scheduled daily publication.
Outcomes are matched to the exact completed target month with the original series identifier, units and geography. Absolute error = |actual − prediction|. The benchmark always predicts the last value available at issue. Current scoring uses the latest retrieved history, so revised source observations can revise scores. The export includes that scoring source alongside the unchanged original edition.
Unavailable models stay visible. Historical error ranges are not calibrated probability intervals. No project budgets or supplier documents are stored in this public archive. The page lists editions from the 100 most recent stored records; an older edition can still be exported by its exact key.
Edition fingerprint and model version
009e4ed0ea3ae9f46954342413f2f0128afe54ce076e5c39d1ce33c74cf87da8
construction-forecast-v1
Your project inputs stay in this browser unless you explicitly export or share them. Public source requests do not include your budgets.