# H07_E_EXPLORATORY_FREEZE_v0_1

**Created:** 2026-09-13T23:23:41Z  
**Artifact:** `H07_E_EXPLORATORY_FREEZE_v0_1.json`  
**sha256:** `175d4c1a414e3dae9e6df355e18c6f3687511c3a275ea2f1baadaa8adfc22ec5`  
**Hub job:** `77356229-193e-4693-a90c-cae8dcb0bc4a`  
**James:** post-H01-KILL / H04-H05-MECHANICAL / H06-KILL — DIFFERENT cheap freeze→run (WPS product supplied→STEO demand); no inventory redo; no weather redo; no H01/FT900/ESR/AAR

## STATUS
| Field | Value |
|-------|-------|
| STATUS | EXPLORATORY |
| VINTAGE_STATUS | UNVERIFIED/INCOMPLETE |
| CLAIM_STRENGTH | LOW |
| CAPITAL_USE | PROHIBITED |
| MODEL_CHANGE | NO |
| REAL_CAPITAL | PROHIBITED |
| TARGETS_OPENED_AT_FREEZE | NO |

## Why runnable today
EIA weekly `WGFUPUS2` + `WDIUPUS2` hist xls HTTP 200 (release stated 9/10/2026) + **141/141** STEO `*_base.xlsx` already on disk. High-frequency product-supplied pulse → official STEO gasoline+distillate consumption outlook lag. Avoids inventory-copy class, weather class, and blocked archive lanes.

## Locked (all 10)

### 1. Eligible reconstructed sample
- **140** adjacent calendar-consecutive STEO pairs (jan2015→sep2026)
- Forecast dates: mix of PARSED + SCHEDULE_PROXY (same issue clock as H04/H05/H06)
- Every obs **VINTAGE_UNVERIFIED**

### 2. Predictor
`eia_wps_gas_distillate_ps_surprise_mean_inter_issue` — mean combined (gas+distillate kb/d) surprise vs training calendar-month norms.  
Release proxy = week_end (Friday) + 5d (SCHEDULE_PROXY).

### 3. Predictor window
`origin_forecast_date < release_proxy <= destination_forecast_date`.

### 4. Normalization
Training-only (destination years ≤2019). No holdout peek.

### 5. Target
`demand_dest[M] − demand_origin[M]` with **MGTCPUSX + DFTCPUS** (4atab), M = destination year-month. **Values closed at freeze.**

### 6. Baseline (Model A)
Intercept + destination-month seasonality + prior revision (+ missing flag).

### 7. Chronological split
- Development destination years **2015–2019** (N=59)
- Holdout destination years **2020–2026** (N=81)

### 8. Metric
**RMSE** primary; MAE + directional accuracy secondary.

### 9. Success threshold
**≥ 5.0%** relative holdout RMSE improvement.  
**KILL** (<1% or neg / N<8) · **HOLD** ([1%,5%)) · **EXPLORATORY_PROMOTE** (≥5%; NOT capital)

### 10. Trial budget
ONE spec: Model A baseline; Model B = A + one WPS combined-PS feature. No rescue.

## Expected sign
**POSITIVE** (product-supplied upside surprise → upward gasoline+distillate consumption revision).

## Mechanical risks (if promote)
STEO may absorb WPS/PSM product-supplied into near-history; live hist xls not first-print; Wednesday release proxy; COVID holdout shock.

## Next
1. Post freeze to Hub (/docs + comment + /reply chatgpt+all). Never /james.
2. Only then open demand cells and run once.