Product
Backtest a trading strategy before you trade it.
US stocks and ETFs. Put the rules into plain language, or set them yourself. ATLAS simulates the path on historical bars. Paper trading uses virtual capital. Live brokerage execution is coming. It is not available today.
How you work a setup
You do not need to write Python for Builder or SmartLab modes. Say the rules in chat and let Orbit populate the fields, or click them in yourself. Then run the backtest and read the result.
One concrete example
Long-only on SPY, daily bars: enter when the 14-period RSI finishes a closed bar below 30, and exit with a 7% take-profit. Leave slippage and commission at zero until you decide to stress the path. Compare the run to buy and hold.
You can specify that setup today. It is not a backtest result, and it is not a claim that this rule works.
What ATLAS actually simulates
ATLAS loads historical open, high, low, close, and volume from Alpaca. A signal is judged on a closed bar. The simulated fill is the open of the next bar. Slippage and commission start at zero until you change them. The default comparison is buy and hold. You can pick one benchmark ticker. VIX is rejected.
Backtesting step by step
The AI backtesting page is the short version of the whole workflow. The backtesting guide shows how to define a rule, add cost assumptions, choose a comparison, and read the simulation before moving to paper trading.
Markets and tools in scope
US equities and ETFs, plus major crypto pairs such as BTC/USD and ETH/USD. Indicators available today: RSI, EMA, SMA, MACD, Bollinger Bands, ATR, ADX, Stochastic, VWAP, Momentum, CCI, and Williams %R.
After the simulation
Paper trading uses virtual capital. You can place manual market, limit, stop, and stop-limit orders, or arm a paper deployment that watches a saved strategy. Alerts email you once when an entry condition fires, then turn off. They do not submit orders. A finished backtest can be opened from a read-only share link without an account.
SmartLab
Logistic regression, random forest, and XGBoost can train inside a backtest, including walk-forward. Live alerts and paper deployments do not run those models. Fundamentals from Finnhub and FMP are research context. They cannot be entry rules.
Limits
Stratos is not a broker and not an advisor. It does not give personalized buy, sell, or hold recommendations. Options, futures, and forex are out. Backtests are historical simulations under stated assumptions. They are not a record of live trading and not a promise of future results. See the Terms of Service for the hypothetical-performance disclosure.