AI-powered strategy backtesting

Backtest any trading idea in plain English.

Describe your strategy in a sentence. Stratfy parses it with AI, simulates it against a price file you bring, and stress-tests it across 1,000 Monte Carlo scenarios — all in your browser, with your data never leaving it.

Start free → See how it works
No account · No credits · Bring your own AI key
No key yet? Get one from Anthropic, OpenAI or Google.
What Stratfy is

Trading strategy backtesting software that runs in your browser.

Stratfy is a browser-based trading strategy backtester. You describe a strategy in plain English — for example, "buy the 9/21 EMA cross when RSI is above 50, exit at RSI 70 with a 2% stop" — and an AI model converts that sentence into explicit entry, exit, stop-loss and filter rules. Those rules are then simulated against daily OHLCV price data you supply as a CSV for US equities, ETFs or Brazilian B3 tickers, with per-side commission and a modelled slippage assumption applied, and stress-tested across 1,000 Monte Carlo resamples.

It is built for discretionary and systematic traders who want to validate an idea without writing code. There is nothing to install and no spreadsheet to maintain. You paste your own Claude, OpenAI or Gemini API key, so usage is unlimited and your provider bills you directly at cost.

Illustrative preview

Backtest your strategy against real market data.

The figures and chart below are illustrative, not the result of a real backtest.

stratfy.io/app/
Stratfy
Backtest Monte CarloOptionsCompare
∞ BYO key
Total Return
+148.2%
Sharpe
1.84
Max DD
−12.4%
CAGR
+13.9%
Equity Curve — illustrative
Keep scrolling to try the demo ↓
Monte Carlo · 1,000 paths
5th pctile+38.2%
Median+112%
95th pctile+247%
Risk of ruin0.4%
Win rate
Hit rate61.4%
Avg win2.1R
Profit factor1.94
Trades214
Strategy parsed by AI
The pipeline

How AI strategy backtesting works.

You type

Buy when the 9-day moving average crosses above the 21-day one and the 14-period RSI is above 50. Add to the position on pullbacks to the 9-day average, but only while the trend is up. Exit when RSI pushes past 70, and always cap the loss with a 2% stop while taking profit at 4%.

AI parses

Rules out

entry: EMA(9) × EMA(21)
filter: RSI(14) > 50
stop 2% · target 4%
exit: RSI ≥ 70
The lab

Everything you need to test a trading strategy.

AI Strategy Parse

Plain English in, precise rules out. Entries, exits, stops and filters extracted automatically by AI.

Real-Data Backtests

Simulated on the prices you load, with per-side commission and a modelled slippage assumption — not synthetic curves. Order size is not modelled, so costs are a floor; the Options Lab additionally models an assumed bid-ask spread.

Monte Carlo

1,000 futures resampled from your trades with replacement. Ruin probability, confidence bands and drawdown distributions.

Options Lab

Payoff diagrams and Greeks for multi-leg structures, priced with Black-Scholes. The historical backtest sits behind them as a secondary, assumption-heavy view.

Compare Strategies

Run saved strategies side by side over the same date range and starting capital. Each keeps its own commission setting.

Python Export

A readable Python translation of your strategy rules, to inspect and build on. Not a trading bot: it places no orders.

Specifications

Capabilities at a glance.

Stratfy technical specification
CategoryTrading strategy backtesting software
PlatformBrowser-based. No install, no local Python environment.
Strategy inputPlain English sentence, parsed into rules by AI. No coding required.
Markets coveredUS equities and ETFs, major international tickers, and Brazilian B3 equities, FIIs and BDRs (.SA suffix)
IndicatorsSMA, EMA, RSI, MACD, Bollinger Bands, ATR, Stochastic, VWAP
Risk-free rateUS T-bill for USD strategies; CDI / Selic for BRL strategies
Monte Carlo1,000 resampled paths per run, with ruin probability, percentile bands and drawdown distribution
Portfolio modeMulti-ticker basket, up to 20 tickers per strategy
OptionsBlack-Scholes pricing, full Greeks, multi-leg structures, rolling expirations. European pricing applied to American contracts, one flat volatility across strikes, realized volatility as an IV proxy, dividends as a trailing continuous yield, and a modelled bid-ask spread: the payoff and Greeks are exact for the structure you define, the historical P&L is an approximation
Costs modelledEquity: commission in basis points per side, plus a modelled slippage of bps + a fraction of ATR per side — grounded in shape, a stated assumption in size, and blind to order size, so it is a floor. Options: commission plus a modelled bid-ask spread that widens out of the money — also a stated assumption, not quoted data
Export formatsPython translation of the strategy rules (not a trading bot), CSV trade log, strategy library JSON
LanguagesEnglish and Brazilian Portuguese
PriceFree. No account, no credits, no subscription — you supply your own Claude, OpenAI or Gemini API key and pay that provider directly.
FAQ

Questions, answered.

Do I need to know how to code?

No. You write your strategy as an ordinary sentence and the AI turns it into rules. Stratfy is built specifically for traders who do not write code. If you do code, you can still export the finished strategy as a Python script.

Is Stratfy free?

Yes, entirely. There are no plans, no credits and no subscriptions. You paste your own Claude, OpenAI or Gemini API key and pay that provider directly for the AI calls — typically a fraction of a cent per strategy parse. Backtests, Monte Carlo runs and charts execute in your browser and cost nothing.

Do I need an account?

No. There is no sign-up, no password and no email required. You paste your AI API key when you open Stratfy; it is held in memory for that session only and never written to storage, so you paste it again after a reload. Saved strategies live in your browser.

Does it support Brazilian stocks, B3 and FIIs?

Yes. Stratfy supports B3 equities, FIIs and BDRs using the .SA ticker suffix — for example PETR4.SA, VALE3.SA or HGLG11.SA. Brazilian strategies automatically use CDI (or Selic, if CDI is unavailable) as the risk-free rate instead of US T-bills. The benchmark stays SPY unless you set it to IBOV (^BVSP).

What data does it use?

Whatever you load. Stratfy does not fetch market data: you bring your own daily OHLCV file as a CSV, for any ticker you have data for, and it is read in your browser and never uploaded. Columns are matched by name in English or Portuguese, comma or semicolon separated, and a file with an adjusted-close column is adjusted for you. Every simulation applies commission in basis points per side. The equity backtest charges commission plus a modelled slippage of a few basis points plus a fraction of ATR per side, moved against the trade on both entry and exit. Its shape is grounded, but its size is an assumption rather than quoted data, so the app shows what half and double it would cost. Order size is not modelled, so the figure is optimistic for a large account. Entries fill at the signal bar's close by default, or at the next bar's open if you choose that. The Options Lab additionally models a bid-ask spread, also as a clearly-labelled assumption.

Is backtesting accurate?

A backtest is an estimate, not a promise. Stratfy charges commission on every fill. The equity backtest charges a modelled slippage as well, but its magnitude is an assumption and order size is not modelled at all, so treat it as optimistic; the Options Lab also models a bid-ask spread, again as an assumption rather than observed quotes. It also runs 1,000 Monte Carlo simulations that resample your trades with replacement, so each path draws a different mix of wins and losses — which shows how much of a result could have come from luck. It cannot eliminate overfitting: a strategy tuned until it looks perfect on past data usually disappoints in future data.

How is this different from TradingView, QuantConnect or Composer?

TradingView needs Pine Script and QuantConnect needs Python or C#; both assume you can code. Composer uses a visual block editor and is limited to US symphonies. Stratfy takes an ordinary English sentence as the input, covers Brazilian B3 tickers alongside US markets, runs entirely in the browser, and is free with no account — you supply your own AI key.

Which AI providers can I use?

Anthropic Claude, OpenAI and Google Gemini keys all work. The key is held in memory for your session only, never written to storage, and sent with each request through our Cloudflare Worker proxy, which forwards it to that provider without logging or storing it. Usage is billed on your own provider account.

Is this financial advice?

No. Stratfy is simulation software, not an advisor or a broker. Results are hypothetical and past performance never guarantees future results. Nothing on this page is a recommendation to buy or sell any security.