SCORECARDARENASDISCOVERYFOUNDATIONSFOR AGENTS

← HOUSING / USVINTAGE 7053e6c1PENDING

New England · 3M

On 4 August 2026 we predicted New England house prices would be 1.37% higher by August 2026.

Our stated range was 451.778 to 472.241 — if the method works as intended, the true figure lands inside it about 95% of the time. Verdict due the first FHFA release covering August 2026, when the official figure is published. It has not been graded yet, and nothing on this page can change what was written above.

01 / WHAT WAS LOCKED

FORECAST · INDEX LEVEL461.896

CHANGE FROM BASE+1.37%

95% INTERVAL451.778 … 472.241

BASE — LAST FIGURE PUBLISHED BEFORE THE LOCK455.650 · 2026-05-01

TARGET PERIOD2026-08-01

LOCKED AT — TIMESTAMP ASSIGNED BY THE DATABASE2026-08-04T22:05:04.17629+00:00

02 / WHEN THIS GETS ITS VERDICT

This forecast is graded against the FIRST figure the publisher releases for August 2026 — the number the market actually saw — and never against a later revision. That rule was fixed on 2026-08-03, before this forecast existed.

Headline grade (first_print): the earliest publisher vintage at or after the target period. Available roughly two months after the period ends, on the publisher's schedule, not ours.

Companion grade (settled_t_plus_12): the value as published twelve months after the target period, which measures how much the first figure was revised. Available roughly fourteen months after. Both definitions were committed before any grade existed under either.

INSUFFICIENT DATA

This forecast has not been graded. It cannot be: the period it names has not been published yet. When it is, the verdict appears here — pass or miss — and nothing on this page changes.

03 / PROVENANCE

VINTAGE ID7053e6c1-2972-4f52-8923-e2c778e58227

ARENAhousing_us

SERIESNEW_ENGLAND

TARGET PERIOD2026-08-01

LOCKED AT2026-08-04T22:05:04.17629+00:00

MODELar12_logdiff

MODEL SPEC1.1

MODEL SELECTION425 observations >= 60

POINT STATISTICmedian

FIT WINDOW425 observations, 1991-01-01 → 2026-05-01

ARCHIVE AS OF2026-08-04

COMMIT SHA63e0fa12edfd75bea8ff26683a080669bd5c9473

RUN MANIFEST9d992102-c049-4495-9a2f-1cfe4ca48d68

PUBLISHER VINTAGES THIS FORECAST WAS BUILT FROM

Every input the reconstruction used, with the time the publisher released it. The forward-lock check reads these back from the record and fails the run if any of them postdates the lock above — that is what makes "locked before the outcome existed" a checked statement rather than a claim.

INPUT	PUBLISHER RELEASE TIME
observations:NEW_ENGLAND@2026-08-04	2026-08-04T00:00:00Z

CONTENT DIGEST

f493b21d8342a955be525044f1347000bf6345440fcf9b768bb1049121b0dc36

SHA-256 over the payload serialized with object keys sorted lexicographically at every depth and no whitespace, computed when this page was built. It is not stored on the row, and it is not the commitment: the commitment is the append-only row and the lock timestamp the database assigned to it. Fetch the same payload from /v1/vintages and repeat this to check it.

04 / THE LOCKED PAYLOAD, VERBATIM

{
  "as_of": "2026-08-04",
  "model": "ar12_logdiff",
  "interval": {
    "label": "95%",
    "lower": 451.7777885424855,
    "upper": 472.2405729383289,
    "method": "ma_representation",
    "log_variance": 0.0001277068207808182
  },
  "base_value": 455.65,
  "commit_sha": "63e0fa12edfd75bea8ff26683a080669bd5c9473",
  "fit_window": {
    "last_period": "2026-05-01",
    "first_period": "1991-01-01",
    "observations": 425
  },
  "base_period": "2026-05-01",
  "forecast_value": 461.8958775547954,
  "horizon_months": 3,
  "forecast_inputs": [
    {
      "ref": "observations:NEW_ENGLAND@2026-08-04",
      "as_of": "2026-08-04T00:00:00Z"
    }
  ],
  "point_statistic": "median",
  "run_manifest_id": "9d992102-c049-4495-9a2f-1cfe4ca48d68",
  "model_spec_version": "1.1",
  "model_selection_reason": "425 observations >= 60"
}

GET /v1/vintages/7053e6c1-2972-4f52-8923-e2c778e58227

The One Lab · Toronto · Every number on this site resolves to a primary source or a graded model output. Where it can't, we say so.VINTAGE 7053e6c1 · WRITE-ONCE · NOT YET GRADED