Tradezentient is an experimental platform that uses AI to turn trading ideas into testable strategies. Describe your idea in plain English, and our engine generates the code, runs it against historical market data in a secure sandbox, and returns a full performance report — usually in under a minute. The goal is to compress the iteration loop so you can test dozens of hypotheses in the time it would normally take to code one.
What You Can Test
Backtests. Submit a strategy idea — a moving-average crossover, a news-sentiment filter, an earnings drift play, whatever — and get back trades, P&L, hit rate, equity curve, best/worst trades, and an AI-generated plain-language review with a 1–10 score. Default starting capital is $1B (configurable) so position sizing isn't a constraint.
Research mode. Ask questions about market behavior rather than testing a strategy: weekday seasonality, intraday volatility patterns, cross-ticker correlations, news-sentiment effects on next-day returns, earnings surprise drift. You get charts, statistics, methodology, findings, and caveats — like a working session with a quant analyst.
Live mode. Promote any completed backtest to live execution. The strategy runs continuously on streaming minute data during market hours (4:00 AM–8:00 PM ET) and notifies you via web push or SMS when entry/exit conditions trigger.
What Data Is Available
US-listed stocks and ETFs only. No options, futures, FX, crypto, or non-US securities.
You can mix three data sources in any strategy:
- Price. Minute-level OHLCV bars covering 4:00 AM–8:00 PM ET every trading day, including pre- and post-market sessions. Sourced from a licensed Nasdaq data provider.
- News & sentiment. Article-level news with per-ticker sentiment scores (-1 to +1), relevance scores, topic tags, source, and summary. Useful for sentiment overlays, event studies, and headline-reaction strategies.
- Experimental fundamentals & macro. Quarterly earnings (reported vs estimated EPS and revenue, surprise %), 25 fundamental fields per quarter (income statement, balance sheet, cash flow), insider Form 4 transactions, analyst consensus estimates with revision history, and 11 macro series (Fed funds, Treasury yields across the curve, CPI, unemployment, retail sales, real GDP, Brent crude). Marked experimental — schema and refresh cadence may evolve.
History depth scales with your subscription tier, ranging from 6 months on the free plan up to 20 years of price history on the top tier. Full limits are listed on the pricing page.
Important Caveats — Please Read
These matter for interpreting your results honestly:
- Prices are not adjusted for splits or dividends. Multi-year backtests crossing a corporate action will show artificial gaps. If you're testing on long horizons, account for this in your strategy logic.
- Missing minutes are forward-filled with zero volume so strategies see a complete intraday grid. Don't rely on volume thresholds without guarding against synthetic bars.
- Survivorship bias. Coverage of delisted, acquired, or removed companies is not guaranteed and varies by symbol. Long-horizon backtests built around large-cap or index-component universes will skew favorable, since those lists are by definition composed of names that survived and grew.
- News data is sparse and irregular. Articles arrive when they arrive — it's not a continuous time series. Strategies access news as discrete events, not as a feed.
- Pre/post-market liquidity differs from regular hours. Bars from 4:00–9:30 AM and 4:00–8:00 PM ET are included and treated uniformly by the engine, but real-world fills in those windows are harder to achieve than the backtest implies.
- Live mode is not low-latency. Signals are bounded by the 1-minute poller cycle plus a few seconds of plumbing. Not suitable for sub-minute strategies.
- Strategies run with no internet access. This is a security feature — your generated code can't make external API calls, scrape websites, or exfiltrate data. All market data is provided by the engine.
What Tradezentient Cannot Do
- Guarantee profitability, accuracy, or correctness of generated code. AI models make mistakes.
- Provide personalized financial or investment advice. This platform is a research tool.
- Replace careful review of generated logic before any live deployment of capital.
- Account for slippage, fees, borrow costs, or market impact beyond what your strategy code explicitly models.
Important Warnings
Strategies, signals, and analyses produced by Tradezentient are generated by AI and may contain errors, omissions, or look-ahead bias that isn't immediately obvious. Always review the generated code, verify the logic against your intent, sanity-check the results, and consult qualified professionals before risking capital. Past performance in a backtest is not indicative of future results. Information on this platform is not investment advice. Use at your own risk.
All features, data coverage, notification delivery, and platform functionality are subject to our Terms of Service, which contain further details, limitations, and disclaimers.