Vibe Coding #1 - Backtest Trading Strategies with Claude Code | Build Python CLI Tools for AI

Result

Ran 100+ backtests across 7 stocks in minutes - would have taken hours manually.

The problem

Testing trading strategies manually is tedious: run, record, change parameters, repeat
Parameter combinations explode quickly (3 stocks × 5 strategies × multiple params = 100+ runs)
Results end up scattered, hard to compare

The solution

Started from my LinkedIn Learning course materials as reference. The course covers backtesting strategies and optimizing parameters - the heatmap below shows how different parameter combinations affect returns.

Strategy optimization heatmap from the course

The backtesting.py library provides an interactive dashboard showing equity curves, drawdowns, and trade markers for each strategy run.

backtesting.py interactive dashboard

Project setup

Claude Code scaffolded the entire project - CLI with Typer, modular strategy implementations, project-based execution. Each strategy lives in its own module, and a project manager coordinates running experiments across all combinations.

Building the Streamlit dashboard with Claude Code

The key insight

YAML configs define the experiment matrix. You specify symbols, strategies, and parameter ranges - the system expands all combinations and runs them automatically:

# config.yaml
symbols: [AAPL, MSFT, GOOGL]
strategies:
  sma:
    fast_period: [10, 20]
    slow_period: [50, 100]
  rsi:
    oversold: [30]
    overbought: [70]

This config generates 3 stocks × (4 SMA combos + 1 RSI combo) = 15 backtests.

Running experiments

Create a new project with target stocks:

uv run backtest project create portfolio --symbols AAPL,MSFT,GOOGL

Run all strategy combinations across all stocks:

uv run backtest project run portfolio

Simulation results after running backtests

Launch the dashboard to visualize results:

uv run streamlit run frontend/app.py

We tested this on the Magnificent 7 stocks. The Streamlit dashboard displays results as a matrix - which strategy performs best on which stock, with Sharpe ratio, max drawdown, and win rate.

Streamlit dashboard showing strategy performance

Reference

Strategies

Strategy	Parameters	Logic
SMA	fast_period, slow_period	Buy when fast crosses above slow
RSI	rsi_period, oversold, overbought	Buy when RSI < oversold
MACD	fast, slow, signal	Buy when MACD > signal
Bollinger	bb_period, bb_std	Buy at lower band
Breakout	entry_period, exit_period	Buy on N-day high

Project structure

backend

cli.pyTyper CLI

strategiesSMA, RSI, MACD, Bollinger, Breakout

runnerBacktest execution

projectProject management

frontend

app.pyStreamlit dashboard

data/projects/{name}

config.yamlExperiment definition

summary.yamlAggregated results

results/{run_id}Per-run outputs

Resources

LinkedIn Learning Course - reference materials we built upon
backtesting.py - backtesting engine for running strategies
yfinance - fetching historical stock data
uv - Python package manager (fast, handles virtual envs)
Typer - CLI framework for the backtest command
Streamlit - dashboard for visualizing results