Equity curve
Your account value across the whole window, with drawdown shaded underneath rather than exiled to a second chart. One picture, one story: what you made and what you had to sit through to make it.
Product
Everything a backtest owes you
A single return figure is the least useful number in a backtest. These are the parts that decide whether an idea survives a real account — and every one of them is in every report, free tier included.
The input
One sentence is the whole interface
There is no strategy builder, no node graph, no parameters panel. You write what you would say out loud, and that is the strategy:
"Buy when the 20-day average crosses above the 50-day, sell when it crosses back."
A model turns that into a real backtesting.py strategy class. Moving averages, RSI, MACD, Bollinger bands, breakouts, momentum, mean reversion, stop losses and take-profits are all things you can simply name. If a phrase is ambiguous, the generated code shows you which reading it took — which is a far more useful answer than a validation error.
You can also dictate it. The microphone in the composer uses the browser's own speech API, so nothing is uploaded to transcribe it.
The report
Nine things you get back
Every run, every tier. Nothing here is held behind an upgrade.
Plain-language metrics
"Total change", "worst drop", "steadiness", "trades that won". The formal names are one hover away, but the label you read first is the one you already understand.
The full trade log
Every position: date in, date out, price in, price out, percentage and dollar result. A backtest you cannot audit line by line is a backtest you should not act on.
Monthly heatmap
Twelve columns, one row per year. It answers the question a headline return can't: was this steady, or was it one extraordinary quarter dragging four flat years behind it?
Stress tests
The same rules re-run through the February 2020 crash, the 2022 bear market and a long flat stretch. Plenty of strategies look excellent right up until one of those three.
The generated Python
Collapsed by default, one click away, copyable. Run it yourself, change it, or disagree with it. It is evidence rather than a feature.
Compare up to five
Put several versions of an idea on one chart. Runs are aligned on a shared calendar, so a shorter test can never win by accident — a real trap in most comparison tools.
Real costs, real fills
Commission comes out of every trade. Orders fill on the bar after your signal, never the one it fired on. Friction is where most paper strategies quietly die.
Saved strategies
Keep the ideas worth returning to and re-run them against a new window in one click. Your run history sits in the sidebar for as long as the session lasts.
Coverage
What you can point it at
| Market | Examples | History | Resolution |
|---|---|---|---|
| US stocks & ETFs | AAPL, MSFT, NVDA, SPY, QQQ | Back to listing, often 20+ years | Daily bars |
| Major indices | ^GSPC, ^IXIC, ^DJI | Decades | Daily bars |
| Crypto | BTC-USD, ETH-USD, SOL-USD | From listing | Daily bars |
| FX | EURUSD=X, GBPUSD=X, USDJPY=X | Roughly 20 years | Daily bars |
| Commodities | GC=F, CL=F, SI=F | Varies by contract | Daily bars |
Daily bars only, for now. Intraday strategies — anything that depends on what happened between the open and the close — cannot be tested here yet. If that is your idea, this is the wrong tool today, and we would rather you knew that before signing up than after.
Honest comparison
Where this fits
There are good reasons to use something else. Here they are.
| AlphaBack | Python by hand | Charting platforms | A spreadsheet | |
|---|---|---|---|---|
| Time to first result | Seconds | Days to weeks | Hours, plus a scripting language | Hours, and it will be wrong |
| Needs code | No | Yes, plus data plumbing | Yes — a proprietary one | Formulas, which is code wearing a hat |
| Shows its working | Full Python, every run | You wrote it | Usually a black box | Yes, painfully |
| Intraday data | Not yet | If you pay for it | Yes | No |
| Live trading | No — deliberately | If you build it | Yes, via a broker | No |
| Cost to start | Free, no card | Free, minus your time | Usually a subscription | Free |
Stop guessing.
Start measuring.
Five free runs, no card, about five seconds each.