Two unhealthy patterns hide in trade-by-trade size choices:
- Martingale: size grows after losses. "I'm down €200, I'll go bigger to make it back." The single fastest way to turn a normal drawdown into a margin call. Statistically the worst behaviour pattern in retail trading.
- Anti-martingale (confidence cycles): size grows after wins, shrinks after losses. Less catastrophic than martingale but still drift — your edge becomes hostage to your last week's P&L instead of being driven by your declared risk-per-trade.
A disciplined trader's size is uncorrelated with the previous trade's outcome. Each trade gets the size the SETUP demands, not the size the trader's emotional state suggests.
The detector measures the Pearson correlation r between (sign of previous trade's P&L) and (this trade's volume divided by your median volume). |r| < 0.3 = disciplined; r > 0.3 = anti-martingale; r < -0.3 = martingale.
30 trades over 4 weeks. Median volume = 1 lot. After wins: avg size = 0.5 lots After losses: avg size = 2.0 lots Pearson r = -0.62
Severity bands: - Disciplined — the detector stays silent. Your size is driven by your risk-per-trade rule, not by recent outcomes. - Early drift — the pattern is forming. Time to recommit to your fixed-fractional rule before it becomes a habit. - Established pattern — the detector landed; you're sizing emotionally. Cap your max size at 1× median for a week and watch the correlation drop. - Entrenched — correlation values this strong are statistically unambiguous. If the pattern is martingale → freeze your sizing at 1R per trade until you can articulate WHY each entry deserves more. If anti-martingale → set a floor: never trade below your usual size even after a loss.
The Onyx-Engine assigns your sizing to one of these bands; the cutoff thresholds are TradeOnyx-internal calibration.
Martingale is more dangerous than anti-martingale. Both are drift, but martingale has a non-trivial probability of bankruptcy in a normal Monte-Carlo simulation; anti-martingale just means you under-trade your edge. The detector classifies them separately for a reason.
Pair with the Discipline-Scorecard's risk-sizing axis. That axis measures how often your declared loss size held; this detector measures whether your size is driven by emotion. They co-trigger: a low risk-sizing score AND a strongly drift-flagged sizing pattern is the worst-case combo. Fix the sizing rule first, the scorecard follows.
How to read the card:
1. Hero (left) — the correlation coefficient r with sign + verdict word (Martingale risk / Confidence drift). Read the SIGN first: negative is the dangerous one. 2. Body explainer — pattern-specific paragraph. The Martingale variant explains the margin-call risk; the Anti-Martingale variant explains the under-utilised-edge risk. 3. Two MicroStats — Avg size after wins, Avg size after losses. Both as multiples of your median (e.g. "0.5×" = half your typical size). Side-by-side comparison is the actionable read: "yep, my losses-trades are 4× my wins-trades, that's the pattern."
The fix is a single rule. Whatever your discipline gap looks like, the cure is the same: pre-decide the size BEFORE looking at recent P&L. Risk-per-trade × capital = size. Always. The detector's correlation falls back to zero in 4-6 weeks of disciplined application — measurably so on the next monthly snapshot.
Tier: Pro. Sits alongside R-Multiple (free, the existence-of-edge question) and Setup-Degradation (pro, the per-strategy fade). This detector is the per-EMOTION fade — what your sizing tells you about your inner state, separate from any single setup or symbol.