Portfolio cleanup: reproducible S&P 500 vs KOSPI analysis

[Jimin0731]

Motivation

• Rebuild the repository as a focused, reproducible portfolio comparison of S&P 500 (^GSPC) and KOSPI (^KS11) local-currency monthly price indices from January 2000 through December 2024.
• Correct the original frequency mismatch, personal paths, sector-average mislabelling, stale notebook outputs, and failed forecasting experiments.
• Keep normal analysis and CI offline after committing a canonical monthly snapshot.

Implemented

• Added canonical analysis.py functions for monthly returns, normalized performance, calendar-year compounding, annualized geometric price return, annualized volatility, maximum drawdown, positive-month ratio, 12-month rolling volatility, and 36-month rolling return correlation.
• Added a validated data refresh script, committed monthly snapshot and metadata, machine-readable results, four deterministic UTF-8 SVG figures, README consistency tooling, a presentation notebook, fresh-kernel notebook smoke execution, focused tests, pinned dependencies, and read-only offline CI.
• Archived the original coursework without modifying the original PDF.
• Fixed CI import handling and PNG-validation false positives.

Data snapshot

• Delivery: live Yahoo Finance through yfinance
• Tickers: ^GSPC and ^KS11
• Requested window: 1999-12-01 through 2025-01-01
• Common snapshot: December 1999 through December 2024
• Snapshot rows: 301
• Reported monthly observations: 300 per market
• Missing values: 0
• SHA-256: d6627cb85439f6dc7724b4d5efbc4ced9ac0ada89f88449bef358bf9a15aeb4f

Key results

Market	Annualized geometric price return	Annualized volatility	Maximum drawdown	Positive months	Best year	Worst year
S&P 500	5.71%	15.30%	-52.56%	62.33%	2013: 29.60%	2008: -38.49%
KOSPI	3.45%	21.16%	-49.18%	54.67%	2005: 53.96%	2000: -50.92%

The 36-month rolling monthly-return correlation has mean 0.6602, minimum 0.2686, and maximum 0.8400.

Interpretation limits

Results use unadjusted local-currency price indices. Dividends, USD/KRW exchange-rate movements, fees, taxes, inflation, and investor-specific currency conversion are excluded. This is descriptive historical analysis, not a forecast, causal claim, recommendation, trading strategy, or total-return comparison.

Validation status

• Generated data, all three result files, README values, and all four SVG figures are committed.
• Snapshot hashes agree across metadata and result files.
• The earlier pytest failure was caused only by the PNG checker scanning its own deny-list string; that self-reference has been fixed in head 38d6d42c8fc86399c459227347d2ed0c934aa980.
• The latest GitHub Actions runs are currently queued/pending for final offline validation.

This PR remains draft and unmerged.