Homework 4

Homework 4#

Note

Homework 4 has been reworked. An earlier version of this assignment was an options case study; that assignment has been retired, and the options notebooks remain class material in Week 7. Homework 4 is now a short, focused assignment: deploy a live, self-updating clone of the CME FedWatch tool.

Overview#

Homework 4 turns the in-class FedWatch case study (finm-32900/case_study_fedwatch) into a live monitor. The pipeline pulls 30-Day Fed Funds futures (ZQ) from Databento and the effective federal funds rate (EFFR) from FRED, computes the market-implied probability of a hike, cut, or no change at the next FOMC meeting, and renders the forecast chart into a chartbook site. You will (1) fill in the small amount of math that has been removed and (2) publish the site so that a GitHub Action rebuilds it every morning, unattended.

The two class notebooks from Week 8 walk through everything the pipeline does:

This assignment is the Week 8 material—GitHub Actions, cron scheduling, and GitHub Pages—put into practice. It is also the course’s themes in miniature: a reproducible analytical pipeline that runs end-to-end with no human in the loop, from raw data pull to published product.

Learning Outcomes#

  • Schedule recurring jobs with cron in GitHub Actions

  • Publish a static site with GitHub Pages

  • Manage API keys with Actions repository secrets (in CI) and .env files (locally)

  • Run a full doit pipeline unattended in CI

  • Understand how fed funds futures prices imply FOMC meeting-outcome probabilities

What You Do#

The assignment repository’s README spells out every step; in brief:

Task 1: Fill in the FedWatch math (1 point)#

Three functions in src/fedwatch.py have had their bodies replaced with raise NotImplementedError(...): implied_rate, solve_post_meeting_rate, and move_probability. Each docstring specifies exactly what the function must do, and the replication notebook derives the same formulas step by step.

Your feedback loop runs offline, with no API key:

pytest -vv ./src/test_fedwatch.py ./src/test_fedwatch_monitor.py

Once those tests pass, run the full pipeline locally: copy .env.example to .env, add your Databento API key (the pull is cost-guarded and free), and run doit. The built site lands in docs/index.html.

Task 2: Deploy your daily self-updating site (2 points)#

The workflow in .github/workflows/deploy_pages.yml rebuilds the pipeline and publishes the chartbook site to GitHub Pages—on every push to main and on a daily cron (14:30 UTC, mid-morning Chicago, after the NY Fed’s ~9 AM ET EFFR print). You will:

  1. Push your completed pipeline to a new public repo under your personal GitHub account (the assignment repo itself stays private).

  2. Add your DATABENTO_API_KEY as an Actions repository secret—never commit it in code.

  3. Run the workflow once by hand and watch it go green: it pulls the data, executes the notebooks, renders the forecast chart, builds the site, runs the tests, and pushes the result to the gh-pages branch.

  4. Enable GitHub Pages on gh-pages and confirm your site is live at https://<you>.github.io/<repo>/.

  5. Back in the assignment repo, record your attestation: set the flag to True and paste your live site URL in src/monitor_self_attestation.py, then commit and push.

From then on, the cron refreshes your forecast every morning with no action from you. I will visit the URL you provide and check that the site is live and current.

Grading#

3 points total, autograded on every push by GitHub Actions:

Component

Points

FedWatch math tests pass (Task 1)

1

Monitor attestation: flag set (Task 2)

1

Monitor attestation: live site URL provided (Task 2)

1

I will spot-check the attested URLs to confirm the sites are live and updating on schedule.

Warnings#

  • The assignment repo must stay private—it is your graded work. Your separate deploy repo will be public, including your completed Task 1 code; that is intended for this assignment.

  • Your API key goes into .env locally and into the Actions secret in CI. If it ends up in a commit anywhere, revoke it and generate a new one.

  • Do not edit the test files (test_*.py).

Contents