Launching soon — TradeOnyx is still in preparation; public sign-up isn't open yet.

Onyx Patterns

Symbol persistence

Watches each symbol's per-trade P&L across three 30-day windows. Flags symbols whose edge has quietly faded — positive in older windows, negative or much smaller now. Catches the EURUSD that used to print money 60 days ago.

What it is

Some setups work for years; others depend on a market regime that quietly ends. The trader who runs five symbols on autopilot rarely notices when one of them has stopped paying — the broken symbol still produces trades, the winners-and-losers blur into the overall P&L, and only after weeks do they realise it was always the same symbol bleeding.

Symbol persistence runs a per-symbol three-window comparison: 0-30 days ago, 30-60 days ago, 60-90 days ago. For each symbol with at least 3 trades in EACH window, it computes the average per-trade P&L per window and flags symbols where: - recent window's average dropped below 50 % of the prior windows' mean, AND - the trajectory is monotonic-ish (recent < mid; not a one-window blip)

A symbol that lost money throughout (negative prior average) is NOT flagged here — it's a bad symbol, not a fading one. Different problem, surfaced by other detectors (Symbol-Concentration is the cousin that catches consistently-bad symbols).

Formula
Approach: per symbol, the Onyx-Engine compares three rolling time-windows of trades — far, middle, and recent — and computes a fade ratio between the recent window and the prior baseline. A symbol is flagged when (a) the recent window has materially weaker per-trade P&L than the baseline AND (b) the trajectory across the three windows is monotonically worse — not a single bad month bouncing back.
 
TradeOnyx-internal: the window lengths, the fade-ratio cutoffs that separate Mild / Strong / Edge-gone, and the minimum sample sizes per window are calibrated empirically and not published.
Example

EURUSD over 90 days: Far (60-90d): 5 trades, avg +€30 per trade Mid (30-60d): 5 trades, avg +€30 per trade Recent (0-30d): 5 trades, avg -€10 per trade

Result→ Flagged with **Edge gone** verdict. The recent window flipped negative against a strongly-positive prior baseline, and the trajectory worsened across all three windows.
How to read it

Verdict bands — driven by the most-faded symbol's severity: - Mild fade — the symbol is still profitable, just less so. Watch it for two more weeks; could be one bad cluster. - Strong fade — recent average is barely positive against historical positives. Cut the size, don't cut the symbol yet. - Edge gone — recent window is negative against positive prior. Stop trading the symbol until you can articulate WHY it broke OR a fresh window proves it was a blip.

The Onyx-Engine assigns each flagged symbol to one of these bands; the cutoff thresholds are TradeOnyx-internal calibration.

The trajectory matters. A symbol where the mid window is the bad one but the recent window recovered isn't flagged — that's noise, not fade. The detector requires the trajectory to be monotonically worse so a single bad month doesn't trigger.

Pair with Market-Regime detector (TRA-226). A symbol fading often correlates with a regime change — bull-trend EURUSD during a Fed-pivot becomes range-bound EURUSD during a Fed-pause. The regime detector tells you the macro reason; this detector tells you which of your symbols got caught.

Where TradeOnyx uses it

How to read the card:

1. Hero (left) — total fading symbols + verdict word + the worst-affected symbol name. 2. Window label — confirms the three-window comparison sequence (Far → Mid → Recent). 3. Per-symbol row — three at most. Each shows: symbol name, fade-ratio as a red badge, then the trajectory as `Far → Mid → Recent` per-trade P&L numbers (recent in red), plus the trade-count split (e.g. `5/5/5`).

The arrow chain is the visual narrative. Read left → right: Far +30 → Mid +30 → Recent -10. Three numbers tell the whole story: edge held, edge held, edge broke. The trajectory is what separates 'fade' from 'noise'.

Weekly-review pairing: open this card → spot the worst-faded symbol → open the Charts tab for that symbol over the last 90 days. Look for: regime change (trend → range), volatility collapse, or a structural break. If you can articulate the cause, freeze + monitor. If you can't, freeze + audit.

Tier: Pro. Pairs with R-Multiple (free, the headline question 'do you have an edge?') as the per-symbol drill-down.

Related reading