You found a strategy idea — an SMA cross, a breakout, a mean-reversion fade. The fastest way to lose it is to spend three weekends building your own backtesting engine before you ever test it. Here's the checklist I'd follow to go from idea to honest numbers in one evening, with free tools only.

1. Get free data before you write any strategy code

  • Stocks/ETFs: yfinance (pip install yfinance) gives daily OHLCV in two lines.
  • Crypto: public OHLCV endpoints on major exchanges via ccxt — no API key needed for market data on most of them.

2. Clean the data first, always

Duplicate rows, missing days, zero-volume ghosts and mixed timezones will silently wreck your results. Drop duplicates on the timestamp, don't forward-fill holes (a hole usually means the market was closed), and sanity-check that high ≥ low and open/close sit inside the day's range.

3. Don't build a framework — use one

backtrader or vectorbt give you order handling, commission math and performance metrics for free. Your edge is the strategy, not the engine. This is the #1 beginner trap: three weekends of engine-building, zero strategies actually tested.

4. Write the strategy as two rules

Entry condition. Exit condition. If you can't state each in one sentence, the strategy isn't defined yet. Backtest exactly those — nothing else.

5. Include fees and slippage, every time

A 0.1% round-trip cost turns most "profitable" intraday strategies deeply negative. Set the commission in your backtester before you look at any result.

6. Check for look-ahead bias

Nothing in your signal may use future data: a signal computed on today's close should execute at the next bar's open. If your equity curve looks too good to be true, this is usually why.

7. Report Sharpe, max drawdown and win rate — not just total return

A 200% return with a 60% drawdown is not something you can actually hold. Total return alone hides the pain.

8. Walk-forward before trusting anything

Optimize on the first slice of history, then test untouched on the last slice. If the edge dies out-of-sample, it was never an edge — and better to learn that for $0 than with real money.

If you'd rather start from a working skeleton than assemble the pieces yourself, I put together a Python Backtest Toolkit: backtrader template (one command, prints Sharpe / max drawdown / win rate), 3 sample strategies (SMA cross, RSI mean reversion, Donchian breakout), a market-data cleaner for messy CSV/Excel files, and a 500-day sample OHLCV set. $1 launch price, instant download, runs 100% locally — your data never leaves your machine: https://datafixtools.sell.app/product/python-backtest-toolkit

Not investment advice — educational toolkit only. Sample parameters are teaching starting points, not optimized results. No profit promises.