Strategy Transparency

Full breakdown of the NeroVibes signal engine architecture. No black boxes — see exactly how signals are generated.

Signal Generation Pipeline

Market Scan→

Trend Detection→

TQ Score→

Trade Guard→

Regime Filter→

Position Size→

Signal Output

Each signal passes through all 6 layers before reaching users. ~42% of initial detections are filtered out by the guard system.

Signal Source: Trend Pullback (S10)

The sole alpha source. Detects EMA200 + EMA50 trend structure with RSI pullback confirmation on bullish setups.

•EMA200 as primary trend filter

•EMA50 for intermediate momentum

•RSI pullback into oversold zones

•Bullish structure confirmation required

•4-hour timeframe for all pairs

Quality Filter: TQ Score (5-Component)

Trade Quality score with rolling percentile normalization over a 120-bar window. Range: 0-1, mapped to 0.8x-1.2x sizing.

•Trend strength (0.25w): dist_ema200 rolling rank

•Regime clarity (0.20w): |ema_gap| rolling rank

•Pullback quality (0.25w): 1 - RSI rolling rank

•Extension risk (-0.15w): price z-score rank (penalty)

•Vol instability (-0.15w): extreme ATR rank penalty

Risk Guard: Trade Guard (3-Layer)

OR-based veto system. Any single layer detecting risk can block a signal from executing.

•Layer 1: OBV Impulse — divergence detection blocks counter-trend

•Layer 2: Reversal ML (LightGBM) — 20 features, P(reversal) > 55% blocks

•Layer 3: HMM Regime — 3-state (TREND/RANGE/TRANSITION), TRANSITION blocks all

Position Sizing: TQ × Vol

2-layer bounded composite sizing. Prevents oversizing during volatile conditions and undersizing on quality setups.

•TQ Score: 0.8 + 0.4 × tq_score → range 0.8x-1.2x

•Vol safety: LOW/NORMAL = 1.0, HIGH = 0.8, CRISIS = 0.5

•Final bounded: 0.4x minimum, 1.2x maximum

•Base risk: 1% per stop-loss hit

Stop Loss: ATR-Based Dynamic

Each coin gets a stop-loss proportional to its own volatility. No fixed percentage — adapts to market conditions.

•Formula: ATR × 2.5 per coin

•Bounded range: 3% minimum, 12% maximum

•Fallback: 8% if ATR unavailable

•TP levels: 0.4R / 0.8R / 1.3R (partial profit-taking)

Safety: Behavioral Guards

Multiple behavioral guards prevent cascading losses and protect the portfolio from drawdown spirals.

•v7 Direction loss guard: 3 same-dir losses → skip 2

•v7 Portfolio loss streak: 3 any losses → skip 2

•Equity DD governor: DD > 12% → 0.5x position size

•Consecutive loss escalation: 5-6 = 75% size reduction, 7+ = FULL STOP

•Margin cap: 70% maximum utilization

Market Regime Classification

BULL

ADX > 25, EMA50 > EMA200, uptrend confirmed

Favors LONG signals, reduced SHORT

BEAR

ADX > 25, EMA50 < EMA200, downtrend confirmed

Favors SHORT signals, tighter stops

RANGE

ADX < 25, no clear trend direction

Higher confidence threshold, fewer signals

TRANSITION

ADX 15-25, regime shifting

15-25% confidence penalty, 2-bar exit confirmation

Design Principles

Single Alpha Source

One well-understood trend pullback strategy. No ensemble of conflicting models. Clarity over complexity.

Defense in Depth

6 independent risk layers. Each can independently block a bad trade. Redundancy protects against any single failure.

Bounded Sizing

Position sizes bounded 0.4x-1.2x. No extreme bets. 1% risk per stop-loss hit. Equity drawdown governor at 12%.

This page describes the actual production signal engine architecture. Strategy details are provided for transparency. Past performance is not indicative of future results.