my-stocks-experiment

Stock Experiment

A Python implementation of the Nasdaq-100 tech quant trading strategy documented in stock-trading-strategy.md. The strategy scores a static universe of Nasdaq-100 tech/communication-services names on trend, fundamentals, and news-risk pillars, builds a risk-managed target portfolio, and can either backtest that logic or generate a live (no-order-execution) signal report.

Setup

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Market data (prices + fundamentals) comes from Yahoo Finance via yfinance and is cached on disk under .cache/market_data/ (gitignored) to avoid re-downloading on every run.

Running

Everything goes through trading_strategy.py. It has three mutually exclusive modes:

python trading_strategy.py --list-universe   # print the tracked tickers + benchmark
python trading_strategy.py --test-history    # backtest vs. buy-and-hold QQQ
python trading_strategy.py --run             # today's regime/scores/target portfolio (no trades placed)

Note: trading_strategy.py depends on packages installed inside .venv (numpy, yfinance, etc.). Running it with a bare python/python3 from a shell where the venv isn't activated will fail with ModuleNotFoundError. Either activate the venv first (source .venv/bin/activate) or use the run.sh wrapper below, which activates it for you (and creates it on first use if it doesn't exist yet).

./run.sh --list-universe
./run.sh --test-history --quarters 20 --capital 25000
./run.sh --run --holdings my_holdings.json --output today.json

Examples

python trading_strategy.py --list-universe
python trading_strategy.py --test-history
python trading_strategy.py --test-history --quarters 20 --capital 25000
python trading_strategy.py --run
python trading_strategy.py --run --holdings my_holdings.json --output today.json

Flags

Flag	Description
--test-history	Backtest the strategy over N quarters (default 50) vs. buy-and-hold QQQ.
--run	Generate today's regime/scores/target portfolio (live, no trades placed).
--list-universe	Print the tracked ticker universe and exit.
--quarters N	Backtest window length in quarters (default: 50).
--capital N	Starting capital in dollars (default: 10000).
--rebalance {weekly,daily}	How often to fully re-score and rebalance in the backtest (default: weekly).
--no-fundamentals	Disable the fundamental score pillar.
--no-news	Disable the live news overlay in --run mode.
--holdings FILE	JSON file {ticker: dollar_value} of current holdings, for --run.
--output FILE	Write a JSON/CSV report to this path.
--cache-dir DIR	Data cache directory (default: .cache/market_data).
--refresh-cache	Ignore cached data and re-download.
-v, --verbose	Enable debug logging.

Run python trading_strategy.py --help at any time for the full, up-to-date list.

There is no test suite or linter configured yet — validate changes by running --test-history (and --list-universe for quick sanity checks) and reading the printed report.

Architecture

• trading_strategy.py — CLI entry point; argument parsing and the two top-level commands (--test-history, --run) plus report printing.
• strategy/universe.py — static ticker universe and sector map.
• strategy/data.py — yfinance-backed price/fundamentals fetch with an on-disk parquet cache.
• strategy/indicators.py — plain price-based technical indicators (SMA, ROC, ATR, percentile rank), no external TA library.
• strategy/signals.py — the trend/fundamental/news scoring pillars and the composite score.
• strategy/portfolio.py — regime detection, position sizing/caps, and stop-loss/drawdown rules.
• strategy/backtest.py — the backtest engine.
• strategy/live.py — --run mode: computes today's regime/scores/target portfolio and a trade list to get there from given holdings. Deliberately has no order-execution code.

See CLAUDE.md for more detail on the module layout, and stock-trading-strategy.md for the design doc this code implements.