The Patterns tab is the deterministic Layer 1 of TradeOnyx's analysis stack. While the AI-augmented briefing and weekly digest produce narrative insights, Patterns produces structured findings — and only when the signal is strong enough to act on. Six detectors run on every closed trade in your journal:
- Weekday performance — which days of the week pay off, which bleed.
- Time of day — three UTC session buckets compared head-to-head.
- Discipline after losses — does your win-rate drop after a losing streak? The detector that exposes the gap most traders feel but can't quantify.
- Exit quality — using MFE / MAE on every trade, are you closing winners too early or holding losers too long?
- Symbol concentration — when one or two tickers carry most of your P&L, your overall edge is fragile.
- Setup performance — which tagged setups (your playbooks) actually pay off, and which ones you should retire.
Each detector follows a meaningful-only guard: the card renders only if the signal is real (e.g. ≥10pp spread for the time-of-day detector, ≥5 trades per bucket). A flat distribution returns nothing — silence is the strongest signal that there's no pattern to act on yet.
The page also gates on trade volume: under 20 closed trades, the detectors return `ready=False` and a friendly prompt asks you to keep journaling. Above that floor, every detector that has enough data renders its card.
Reading the cards: - Discipline after losses — fresh win-rate vs. after-N-losses win-rate, with the percentage-point drop quantified. A 10pp drop is the typical threshold for a real pattern; 20pp+ means you're trading materially worse after defeats. The fix isn't a new strategy — it's a circuit breaker ("after 2 reds I'm done for the day"). - Exit quality — winner-capture below 60% means you're consistently closing before the move is done; loser-ride above 70% means you're hoping past your stop. Both are journal-worthy on the next trade. - Setup performance — green tags scale up, red tags get retired. The catch: you need ≥5 trades per setup before the detector trusts the result. Don't kill a setup off 3 trades. - Symbol concentration — if Top-3 symbols carry > 80% of your total P&L, you're not running a strategy, you're running a position in 3 names. Diversification or explicit acknowledgement that you're a specialist — both fine, just be honest about it. - No card rendered — that detector found no signal worth surfacing. It's not a bug, it's the meaningful-only guard. Some detectors will never render for some traders (a swing trader has no time-of-day pattern by design).
The growth-nudge at the top of the page is intentional: more journal data unlocks more detector classes. Setup-quality analysis only works if you tag your trades with the playbooks you're running. Mood × outcome correlation only works if you log mood scores. The detectors you see today are Phase A; richer journal context unlocks the next ones.
Patterns is the weekly review tab. Open it Sunday evening alongside the Journal and the Discipline scorecard — the three together produce the answer to "what changed this week, and what should I do about it?". The detectors don't tell you to trade more or less; they tell you which behaviour was driving the result so you can act on the cause, not the outcome.
In TradeOnyx, the page calls `/api/v1/patterns` once on load — the detectors are pure functions over your closed trades, so the response is fast (typically < 200 ms even on a thousand-trade journal). The growth-nudge at the top stays visible from your first trade onwards; the detector cards appear once you cross the 20-trade threshold.
The pairing that makes Patterns valuable: a finding here is a hypothesis, not a verdict. The next step is Trades-tab filter by the dimension the detector flagged (e.g. "only Mondays" or "only EURUSD"), then Journal review of those trades. The pattern is data; the explanation is in your own notes. TradeOnyx is built around that loop — Patterns surfaces the data, Journal holds the explanation.