PatternX Classic vs PatternX AI: how each works

Every PatternX account runs in one of two modes, chosen with a single account setting: PatternX Classic, the default, or PatternX AI. Both modes find setups in exactly the same way. AI mode inserts one extra evaluation step between the score and your own settings; detection, exits and execution stay the same. This page walks through both pipelines, compares them line by line and shows where large language models fit in.

How PatternX Classic works

PatternX Classic is a deterministic pipeline: the same candles and the same settings always produce the same result.

1. Market data. For each venue, symbol and timeframe you track, the scanner fetches the latest 200 candles, plus order-book and funding data where the market provides them.
2. 20 detectors. Each rule-based detector checks a fixed window of recent bars for one formation, such as a bull flag or a double bottom. A match needs its breakout on the latest bar, which may still be forming. The rules are covered in how PatternX detects chart patterns.
3. Universal filters. Every match must pass the same market-quality checks: breakout volume of at least 1.2× the 20-bar average, ATR% of at least 0.05, a liquidity score of at least 30, a spread no wider than 0.25%, no strongly opposing order book and, for crypto, at least $3M of 24-hour volume.
4. PatternX Score. Survivors are scored from 0 to 100 on 13 weighted factors. The score becomes the setup's confidence, and anything under 50 is discarded.
5. Your settings. Your minimum confidence (65 by default), the patterns you have enabled and your take-profit and stop-loss settings are applied.
6. Risk engine. Before any order, the risk controls check kill switches, daily quotas, loss limits, reward:risk and position size.
7. Signal or execution. In manual mode, the default, the signal waits for your approval and expires after 15 minutes. In automatic mode it is executed without approval, still through the full risk check. New connections start in paper mode.

How PatternX AI works

AI mode runs the same first four steps. It then passes every setup the engine accepted, at confidence 50 or more, to the context layer: a fixed formula with 8 weighted factors covering trend strength, market structure, volatility regime, volume quality, momentum, confluence, historical similarity and funding sentiment. Each factor is scored from 0 to 100 from market data and PatternX's own trade record, and their weighted mean is the context score.

Two outcomes are possible. The setup is vetoed unless its context score is at least 48 and market structure and momentum are not both weak, meaning structure below 35 together with momentum below 40. If it survives, its confidence is re-weighted using the context score, and a short reasoning note is produced from a template. From there it follows the Classic path: your settings, the risk engine, then a signal or an order. If the AI evaluation ever fails with an error, that setup is simply evaluated as Classic.

Side-by-side comparison

What PatternX AI is not

The name invites assumptions, so it helps to be explicit:

• It does not invent chart patterns. It cannot detect, find or create a pattern. It only sees a setup after the deterministic engine has detected, filtered and scored it.
• It is not machine learning. Nothing is trained and nothing learns from data. It is a fixed, transparent weighted formula computed from market data, with no external calls.
• It does not change your trade. Take-profit, stop-loss, position size, risk checks and the execution path are the same as in Classic.
• It does not read the news. It uses no news, social sentiment, order-book or correlation data. Its only sentiment input is the funding rate on perpetual futures.

Where large language models are used (and where they aren't)

PatternX does use large language models, from OpenAI or Anthropic when configured, but only to write. They draft signal explanations, answer questions in the assistant and the support assistant, and write daily, weekly and monthly reports, education lessons and backtest write-ups.

They are kept off the decision path. No language model detects patterns, scores setups, applies the veto, checks risk, sizes positions or places orders, and the models are instructed not to give buy or sell advice. Even the context layer's reasoning note comes from a template rather than a language model. Put simply, a language model can describe a signal after it exists; it cannot create, approve, veto or execute one.

Which should you use?

Neither mode is a shortcut to better results, and PatternX makes no claim that one outperforms the other. They make different trade-offs.

Classic suits you if you want the simplest model to reason about. One score and one threshold decide what goes forward: every setup at or above your minimum confidence, among the patterns you enabled, moves on to the risk engine. It is available on every plan, including Free.

AI suits you if you want a second, rule-based check on market context before a setup reaches you. AI mode starts from every setup the engine accepts at confidence ≥ 50, not only those above your Classic minimum, and decides which of them reach you. So it can drop a setup Classic would have sent, such as a long breakout while recent swing highs and lows keep falling and RSI stays weak, and it can pass a setup that Classic would have held back at your minimum.

Before switching, weigh these points:

• Trade quota. On Pro, AI mode allows up to 5 trades a day, against 100 in Classic. Elite has no plan limit in either mode.
• Same exits. Neither mode changes stops, targets or position size. AI mode changes which setups reach you, not how a trade is managed.
• Test first. Run the same backtest in both modes (AI-mode backtests need Pro or Elite), then run the mode you choose in paper trading before going live. Backtests use neutral market context, so treat the comparison as a rough guide.

Common misconceptions

• "AI mode spots patterns Classic misses." It can't: both modes run the same 20 detectors on the same data.
• "AI confidence is a win probability." The re-weighted confidence is still a score, not a measured chance of success. See confidence score.
• "AI mode learns over time." The formula and its weights are fixed. Only its inputs change, including PatternX's live win rate on a market as trades close.

Limitations of both modes

Both modes read breakouts on a bar that may still be forming, see only formations that fit each detector's window, and scan only the symbols you track at your plan's interval. A veto is a rule-based judgment, not foresight: the context layer can reject setups that would have worked and keep setups that fail. Read what PatternX can't do before trading either mode with real money, and see the PatternX Classic and PatternX AI feature pages for plan details.