Questions, answered.

AlphaBack is an AI backtester. You describe a trading strategy in plain English — for example, "buy AAPL when the 20-day moving average crosses above the 50-day, sell when RSI goes over 70" — and AlphaBack writes the Python backtesting code, runs it against real historical market data, and returns a full performance report: equity curve, Sharpe ratio, max drawdown, win rate, and a log of every trade. It answers one question: would this strategy actually have made money?

No. That's the whole point. You write your rules the way you'd explain them to a friend, and the AI turns them into working Python. You never have to read or touch the code — but it's always shown next to your results, so nothing is a black box. If you do code, you can copy it, tweak it, and run it locally.

You can explore the app and see how everything works without one. Creating a free account takes about ten seconds (email + password, or just sign in) and unlocks running your own backtests. No credit card is ever asked for.

Usually a few seconds end-to-end: the AI writes the code, the engine downloads fresh price data, simulates every trade bar-by-bar, and computes the stats. Longer date ranges or slower market-data responses can stretch it slightly, but you're never waiting minutes.

Start with a classic so you can sanity-check the output against your intuition:

• Golden cross — "Buy when the 50-day MA crosses above the 200-day, sell on the reverse cross."
• RSI mean reversion — "Buy when RSI drops below 30, sell when it rises above 70."
• Trend filter — "Only hold SPY when price is above its 200-day moving average."

Then change one thing at a time — the ticker, the period, a threshold — and watch how the metrics move. That's how you build real intuition.

Anything you can express as rules on price and indicators: moving-average crossovers, RSI and MACD signals, Bollinger Band strategies, momentum and mean-reversion systems, breakout rules, buy-and-hold with trailing stops, seasonal patterns, and combinations of all of these. If you can describe the entry and exit conditions in a sentence or two, AlphaBack can test it.

Anything Yahoo Finance carries: US stocks (AAPL, TSLA, NVDA…), ETFs (SPY, QQQ…), major indices, forex pairs, and large crypto pairs like BTC-USD and ETH-USD. The app has quick market tabs for stocks, crypto, and forex, plus a ticker search.

Yes. In the run settings you can control starting capital, commission per trade, slippage, stop-loss %, take-profit %, and position size. These matter — a strategy that trades hundreds of times can look great with zero costs and lose money once realistic commissions are applied. AlphaBack applies a sensible commission by default so results aren't fantasy numbers.

Yes — comparison mode lets you line up backtests side by side: overlaid equity curves and a metric-by-metric table, so you can see not just which one earned more, but which one earned it with less pain (smaller drawdowns, steadier curve, better Sharpe).

Up to about 20 years of daily data, depending on the asset (many crypto pairs only go back to the mid-2010s). You pick the start and end dates. Tip: always test through at least one bad year — a strategy that has only ever seen a bull market hasn't really been tested.

From Yahoo Finance, fetched through the open-source yfinance library. Each run pulls the data fresh, so it's always current. You get full OHLCV bars — open, high, low, close, and volume — which is what the simulation trades on.

The engine is backtesting.py, a widely used open-source framework. It walks the data bar by bar: your rules are evaluated only on information available at that moment, orders fill on the next bar, and commissions and slippage are deducted from every trade. That avoids the most common backtesting sin — accidentally letting the strategy peek at the future.

Yes, by construction. Look-ahead bias is when a backtest uses information that wasn't yet available — like today's close to decide this morning's trade. The bar-by-bar engine only ever sees data up to the current bar, and indicator values are computed the same way. It's one of the main reasons we run a real engine instead of a spreadsheet approximation.

Prices come split- and dividend-adjusted from Yahoo Finance, so a stock split won't show up as a fake 50% crash in your results. Dividend income itself isn't added as separate cash flow — so for high-dividend tickers, real-world total returns would be somewhat higher than the backtest shows.

Small differences are normal between any two platforms. The usual causes: different fill assumptions (open of next bar vs close of signal bar), different commission/slippage defaults, slightly different data sources, and different indicator implementations (there are several accepted ways to compute RSI, for instance). What matters is that comparisons within one platform are apples-to-apples — which is exactly what strategy iteration needs.

Always. Every result screen includes the exact Python that produced it. This is deliberate: if you can't inspect the code, you can't trust the numbers. Expand it, read it, copy it with one click.

Yes. The code is standard Python using yfinance and backtesting.py — install those two packages (pip install yfinance backtesting) and it runs locally, unchanged. It's your code; use it, modify it, build on it.

It can happen with ambiguous descriptions — "buy on momentum" means five different things to five traders. Two defenses: first, the generated code is always visible, so you can check the rules it actually implemented; second, being specific helps enormously. "Buy when the 20-day SMA crosses above the 50-day SMA" will be translated correctly every time. If a run looks wrong, rephrase with exact indicator names, periods, and thresholds.

Generated code runs in an isolated Python environment on our servers — never in your browser and never with access to your account data. Your machine only ever receives the results JSON and the code text.

Yes — fully free during beta. Unlimited backtests, all metrics, the full trade log, and the generated code. No credit card, no trial countdown, no surprise paywall mid-run. A paid Pro tier (saved history, portfolio analytics, exports) is planned; see Pricing for what's coming.

The core loop — describe, run, read results, copy code — stays free. Pro ($12/mo when it launches) adds unlimited saved strategy history, a portfolio analytics dashboard, side-by-side comparison views, CSV export, and historical stress tests. Beta users who join the waitlist get 3 months of Pro free.

They're yours. Your prompts are used to generate your code and results — not sold, not published, not used to trade against you. Strategy descriptions pass through an AI provider to generate code, as with any AI product, but your saved strategies live in your account and nobody else's.

The minimum for an account to function: your email, your name if you give one, and the strategies/results you choose to save. Passwords are handled by Supabase Auth and never stored in plain text. We don't run ad trackers.

Yes. Ask from inside the app (or via the chat page) and your account and saved strategies are removed. Nothing lingers in a marketing database — there isn't one.

Not necessarily — and anyone who tells you otherwise is selling something. A backtest shows what your rules would have done in the past. Markets change regimes, edges decay, and live execution is messier than simulation. What a good backtest does give you: evidence your idea isn't obviously broken, a realistic picture of its drawdowns, and a baseline to compare alternatives against. That's far more than a hunch — but it is not a promise.

Overfitting is tuning your rules until they perfectly fit the past — at which point they usually stop working on the future. If you tweak parameters twenty times until the equity curve looks perfect, you haven't found an edge; you've memorized history. Defenses: keep rules simple, test across different periods and tickers, and be suspicious of results that only work with one magic number.

No. AlphaBack is an analysis tool — it tells you how a rule set performed historically, nothing more. It doesn't know your finances and it doesn't recommend trades. Decisions and risk are yours; never trade money you can't afford to lose.

No. It never connects to a brokerage and never touches real money. Everything is simulation on historical data. If you decide to trade a strategy live, that happens entirely outside AlphaBack, with your own broker.