# Read the Game — Portland 2026–27 forecast publication package

Snapshot: October 10, 2026. Independent model v2. Expected wins: 44.606; approximately 45–37. This package freezes the saved report and its reproducible research inputs. It does not introduce a new forecast.

## Publish the report

Upload the contents of `site/` to your chosen directory on readthegame.com, for example `/blazers/2026-27/`. `index.html` is self-contained and can also be opened locally. It includes every forecast slot, team review, sources, distribution chart, searchable schedule, and expandable margin components. No server, API key, build process or third-party JavaScript library is required. Your host must permit the report's inline JavaScript for search and disclosure navigation; the complete report remains readable without it.

`site/data/` provides JSON and CSV for your own front end: full state, summary, 82 games, team ratings, Portland minutes, all-team rotations, roster adjustments and win distribution. Publish this folder if you want readers to download the numerical results.

The other folders are the research archive, not required website assets. You can offer the full ZIP as a download separately. This package has not been deployed to your domain.

## Reproduce the independent baseline

From the extracted `model/` directory:

```
python -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt
python calibrate_independent.py
python build_independent.py
python render_independent.py
```

On Windows, activate with `.venv\Scripts\activate`. All required numerical snapshots are supplied for offline calculation after installing dependencies. No market data enter the independent calculations. The previous state supplies already-parsed schedule geometry and an archive for comparison.

The renderer produces `Trail-Blazers-2026-27-Forecast.html` in `model/`; copy it into your published folder as `index.html`. It also regenerates the Portland game CSV. The separate `site/data/` export is a frozen publication snapshot and must be exported again if you change the model.

## Contents

- `site/`: publication report plus machine-readable tables.
- `model/`: complete original independent source/data bundle, expanded, with exact installed dependency versions.
- `archive/`: original market-based state, immutable forecast history and archived pregame probabilities.
- `research/`: official schedule materials and source register, including all NBA endpoint metadata.
- `conversation-attachments/`: the three supplied Robinhood images and a factual price/order transcription. These are historical context, not inputs to the independent forecast.
- `METHOD-AND-DECISIONS.md`: a reconstruction of the substantive research decisions in this conversation, not a verbatim chat transcript.
- `PUBLICATION-NOTES.md`: important qualifications and attribution boundaries.
- `MANIFEST-SHA256.json`: integrity hashes for all package files except itself.

## Updates after games

The existing postgame watch updates the saved report; it does not have a deployment connection to readthegame.com. Publishing this static snapshot does not automatically update your site. Republish updated report/data files after a forecast update, or connect your own publishing pipeline.

Do not rerun the preseason builder against live completed-game state. It reconstructs the October 10 baseline and would reset results. The model README and report describe the postgame update protocol; a standalone production postgame-fetch-and-deploy service is not included.