About backtester.run
backtester.run makes institutional-grade backtesting accessible without requiring a Python environment, a data subscription, or hours of boilerplate code. Describe your strategy in plain English — get a zipline backtest with Sharpe ratio, walk-forward analysis, and Monte Carlo simulation.
Built by
Franco Galuzzi
Quantitative developer and algorithmic trading specialist building production-grade trading systems across crypto, forex, and equities. I turn strategies into automated systems — from rigorous backtesting and portfolio optimisation to live execution bots running on Alpaca, Bybit, and Binance.
My work is characterised by institutional-quality documentation, statistical rigour, and the post-delivery communication that turns one-time clients into long-term partners. I don't just deliver code — I make sure you understand exactly what you have and how to use it.
Why I built this
Every serious backtesting tool requires either writing Python (zipline, backtrader, QuantConnect), learning a proprietary scripting language (TradingView's Pine Script), or paying for expensive institutional platforms. The setup overhead — environment management, data ingestion, debugging indicator code — consumes time that should go into strategy research.
backtester.run removes that barrier. The platform handles data sourcing, environment setup, and execution — you describe the strategy, we run zipline. The result is the same institutional-grade backtest output I use in my own trading systems, accessible to anyone without a line of code.
How backtester.run works
The platform translates plain-English strategy descriptions into validated backtests through a three-stage pipeline:
- Natural language → validated DSL. Claude Sonnet translates your description into a structured JSON document — a domain-specific language with a fixed whitelist of primitives (indicators, comparison operators, signal combinators, position sizing rules). Fourteen semantic checks validate the DSL before any backtest runs. No user-supplied code ever executes.
- DSL → zipline strategy. The validated DSL is compiled into a zipline
TradingAlgorithm. Zipline enforces strict bar-by-bar execution: signals fire on bar close, fills execute at the next bar's open. This eliminates the most common class of look-ahead bias by design. - Zipline → results. The engine runs against live OHLCV data from Alpaca (US equities) and Bybit (crypto) at daily, hourly, or minute resolution. Output includes Sharpe ratio, Sortino ratio, Calmar ratio, CAGR, maximum drawdown, win rate, profit factor, a full equity curve, and a trade-level log.
The technology stack
| Layer | Technology |
|---|---|
| Backtest engine | Patched zipline-trader (Python 3.11, event-driven) |
| NL translation | Claude Sonnet (Anthropic) |
| Equity data | Alpaca Markets (OHLCV, live feed) |
| Crypto data | Bybit (OHLCV, spot and derivatives) |
| API | FastAPI (Python 3.11), deployed on Fly.io |
| Frontend | Next.js 14 App Router, deployed on Vercel |
| Auth | Clerk |
Statistical rigour by design
The three biggest reasons backtests fail in live trading are look-ahead bias, overfitting, and survivorship bias. backtester.run addresses the first structurally (zipline's execution model makes bar-close future leak impossible), and provides the tooling to address the second: walk-forward analysis, out-of-sample validation, and Monte Carlo simulation.
This is the same analytical rigour I apply to every trading system I build. backtester.run makes it available without requiring you to build the infrastructure yourself.
Open source
The underlying backtesting infrastructure and related tooling is open-source and available on the Lapunga GitHub organisation. The platform-specific API, web app, and DSL translation layer are proprietary.
Frequently Asked Questions
- Who built backtester.run?
- backtester.run was built by Franco Galuzzi (GitHub: Lapunga), a quantitative developer and algorithmic trading specialist. Franco builds production-grade trading systems across crypto, forex, and equities — from rigorous backtesting and portfolio optimisation to live execution bots running on Alpaca, Bybit, and Binance.
- What backtesting engine does backtester.run use?
- backtester.run runs on a patched version of zipline, the event-driven backtesting engine originally built by Quantopian. The patch adds support for crypto assets, additional timeframes, and live OHLCV data from Alpaca (US equities) and Bybit (crypto). Signals fire on bar close; fills execute at the next bar's open — the same execution model used in institutional backtesting.
- Is backtester.run open source?
- The underlying backtesting infrastructure and related tooling is open-source and available on the Lapunga GitHub organisation. The platform-specific API, web app, and DSL translation layer are proprietary.
- How does the plain-English strategy input work?
- You describe your strategy in plain English — indicators, entry conditions, exit rules, position sizing. Claude Sonnet translates the description into a validated DSL (domain-specific language): a structured JSON document using a fixed whitelist of primitives. No user-supplied code ever executes. The DSL is then compiled into a zipline strategy and run against live historical data.
- What assets and markets does backtester.run support?
- backtester.run currently supports US equities (via Alpaca) and crypto markets (via Bybit). Data is sourced live at daily, hourly, and minute timeframes. Forex and additional exchanges are on the roadmap.