AI trading

Search for "AI trading" and you will find everything from research-grade forecasting systems to simple scanners with a chatbot attached. The term has no fixed technical meaning. In practice, the tools sold under it fall somewhere on a spectrum:

• Rule-based systems apply explicit conditions a person wrote, for example "report a breakout when price closes above resistance on above-average volume". The same input always gives the same output.
• Statistical scoring combines several measurable factors with fixed weights into one number, such as a confidence score.
• Machine-learning models fit their parameters to historical data. They can capture relationships nobody wrote down, at the cost of transparency and a real risk of overfitting.
• LLM assistants (large language models) generate text: summaries, explanations, answers. They are good with language, not at forecasting prices.

This hub explains what each of these actually does inside trading software, where each one helps and where it can mislead you. It is written to be useful whether or not you ever use PatternX.

• AI trading vs rule-based trading compares the four approaches on transparency, reproducibility, overfitting and behaviour when market conditions change, with a side-by-side table and guidance on when each one fits.
• AI chart pattern detection looks at how software finds chart patterns, through geometric rules, image recognition or hybrids, and at the failure modes behind most disappointing results: false positives, lookahead, repainting and overfitting.
• How to choose AI chart pattern software gives you a ten-point evaluation checklist, the red flags that should end a trial early, and an honest account of who PatternX does and doesn't suit.

PatternX sits deliberately at the transparent end of the spectrum. PatternX Classic detects 20 chart patterns with fixed, rule-based detectors and rates each setup with a 13-factor PatternX Score. The optional PatternX AI layer, on the Pro and Elite plans, is a fixed 8-factor context evaluation that runs only after a setup has been detected: it can veto the setup or re-weight its confidence, and it never detects patterns itself. It is not a machine-learning model. Large language models are used only to write explanations, reports and answers, never to detect, size or place trades. The Classic vs AI methodology documents exactly what each mode does.