# Gold, Bitcoin, NVIDIA, and the S&P 500: three-year backtest

## Scope
This package compares a buy-and-hold investment in four assets from 2023-01-06 through 2026-01-06, the latest common date in the included source snapshots.

## Instruments
- Gold: XAU/USD spot gold.
- Bitcoin: BTC/USD spot.
- NVIDIA: NVDA adjusted close.
- S&P 500: SPY adjusted close as an investable proxy.

## Method
1. Use the source-adjusted close where corporate actions apply; spot gold and Bitcoin have no corporate actions.
2. Re-index every asset to one calendar-day index.
3. Forward-fill only from the last observed close; never backward-fill.
4. Rebase each series to 100 on the common start date.
5. Compute calendar-day returns, annualized volatility with sqrt(365), CAGR from elapsed calendar time, and maximum drawdown.

## Files
- market_prices_raw.csv — source rows used.
- aligned_prices_daily.csv — calendar-aligned adjusted prices.
- normalized_curves.csv — growth-of-100 curves.
- performance_metrics.csv — return and risk metrics.
- performance_differentials.csv — all six pairwise terminal-return gaps.
- asset_backtest_curves.png / .svg — final figure.
- source_metadata.json — sources, checks, adjustment semantics, and limitations.
- findings.md — interpretation.
- reproduce_analysis.py — local reproduction from included raw CSV.
- SHA256SUMS.txt — file checksums.

## Important limitation
This is historical evidence, not a forecast. It excludes fees, taxes, slippage, and any future scenario assumptions. Source refresh dates differ, so the common endpoint is constrained by SPY.
