Add a moving-average trend-timing backtest

[valuecodes]

What

• Adds a trend-timing simulation — the fourth strategy in the lab — that
  holds 100% stocks while the S&P is above its moving average and steps aside to
  cash (money-market at the fed funds rate) when it falls below. Answers the
  classic "does market timing beat buy-and-hold?" question none of the existing
  pages cover.
• New engine src/portfolio_research_lab/timing.py (run_ma_timing →
  TimingResult), a TimingConfig Pydantic model, and a 3_Trend_Timing.py
  page. Built as a dedicated path-dependent daily loop (like cash_deploy.py)
  because the Strategy protocol only sets fixed initial weights.
• Methodologically honest by design: the signal is lagged one bar (decided
  from the previous close, executed at today's close — no look-ahead), a
  hysteresis band damps whipsaws, and an optional per-switch transaction
  cost (bps) is modelled — timing is exactly where costs bite.
• Simple or exponential average, with identical warm-up masking; time_in_market
  and switch-count surfaced in the UI. README features + structure updated.

Over 2000–2025 (200-day SMA, 10 bps): max drawdown −23.1% vs −56.8% for
buy-and-hold, 72.4% time in market — the expected "less return, much less risk"
result.

How to test

uv run poe check
uv run poe dev   # open the "Trend Timing" page; try a 200-day window, a band, and a cost

Security review

No security-impacting changes.

[Copilot]

Pull request overview

Adds a fourth strategy to Portfolio Research Lab: a moving-average trend-timing backtest that switches between 100% stocks and 100% cash using a one-bar-lagged MA signal, optional hysteresis band, and per-switch transaction cost, plus a Streamlit page to explore and compare it against baseline portfolios.

Changes:

• Introduces a new path-dependent timing engine (run_ma_timing) returning a TimingResult with equity/position series and a switch-event log.
• Adds a TimingConfig Pydantic model (incl. MA kind validation, band, and transaction cost parameters) and exports the new API from the package root.
• Adds a dedicated Streamlit “Trend Timing” page and comprehensive unit tests covering lag, band behavior, costs, warm-up masking, and input validation.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
tests/test_timing.py	Adds unit tests validating timing engine correctness (lag, switching, costs, warm-up, schema, and error cases).
src/portfolio_research_lab/timing.py	Implements the moving-average trend-timing simulation loop and result container.
src/portfolio_research_lab/models.py	Adds TimingConfig and MA kind validation constants.
src/portfolio_research_lab/init.py	Exposes TimingConfig, TimingResult, and run_ma_timing in the public API.
README.md	Updates features + repository structure to include the new strategy and page.
app/pages/3_Trend_Timing.py	Adds the Streamlit UI page for running and comparing the trend-timing backtest.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.