Edges don't break overnight; they drift. A setup that paid 65 % win-rate at 1.5R for two years can quietly slide to 45 % at 0.8R as market regime shifts, more participants spot it, or your own execution loosens. By the time the equity curve is obviously down, you've spent weeks on a strategy that already stopped working.
Setup degradation runs a per-setup rolling-window comparison: last 30 days versus the preceding 60. For each setup that has at least 5 trades in BOTH windows, it computes the delta in win-rate and the delta in average R-Multiple. A setup is flagged when at least one of the two metrics drops past a noise threshold: - win-rate dropped by ≥ 10 percentage points, or - average R-Multiple dropped by ≥ 0.3R.
The top three most-degrading setups are surfaced. The flagged setups are still ON your trade list — the engine doesn't decide for you. It tells you which setups need a re-evaluation conversation with yourself.
Setup 'Breakout' — recent window shows a much lower win-rate and a much weaker avg-R than its prior baseline. The drop is well beyond ordinary variance.
Verdict bands — driven by the worst setup's drift severity: - Mild drift — the setup is wobbling but the sample may still be small. Don't size up on it; watch for two more weeks before deciding. - Moderate drift — meaningful drop. Re-read your last 10 trades on this setup before the next entry. Look for: market regime change, your own execution drift, or a copycat-flooding effect. - Severe drift — freeze the setup. No more new entries until you can articulate WHY it's broken or the next window proves it's noise. Your equity curve will thank you.
The Onyx-Engine assigns the worst setup to one of these bands; the cutoff thresholds are TradeOnyx-internal calibration.
Don't just delete the setup. Loss-makers contain information: WHY did the win-rate drop? Was it a Fed-pivot regime change? Did you start taking marginal entries because you were chasing a comeback? Was the avg-R hit because you started moving stops? Pair the diagnosis with the Discipline-after-losses card and the Mistake-pattern detector — they often co-trigger.
The min-sample rule matters. A setup with very few trades in either window doesn't appear here even if it's down 80 % — that's noise, not signal. The detector deliberately stays silent until both windows hold enough trades. If your favourite setup never appears, you're either not trading it enough OR your tagging discipline is loose (untagged trades are invisible to per-setup analysis).
How to read the card — top to bottom:
1. Hero (left) — total degrading setups out of total analysed, with a verdict word (Mild / Moderate / Severe drift). Read this first; one number says how worried to be. 2. Window label — confirms the recent-vs-prior comparison window. The detector uses the latest close-time as 'now', not wall-clock time, so cached results stay deterministic. 3. Per-setup row — three at most, sorted by severity. Each row shows: setup name (without `tag:` / `playbook:` prefix), win-rate delta as a red badge, prior-vs-recent win-rate side-by-side, and the avg-R progression with the delta highlighted.
The weekly-review pairing: - Open this card → spot the worst setup → open the Trades tab filtered to that setup → read your last 10 trades on it. Walk through them with the Journal. - 80 % of the time the diagnosis is one of: regime change (open Briefing market-temperature), execution drift (open the Discipline scorecard), or stop-loss creep (open the upcoming Stop-Loss Drift detector when it lands).
Tier: Pro. The free tier shows R-Multiple expectancy (the question 'do you have an edge'); Pro adds the detectors that explain WHY the edge is moving — Setup-Degradation is the per-strategy version of that question.