JevTrader: Institutional Multi-Asset Trend-Following & Convex Pyramiding Architecture in Cryptocurrency Markets

1. Abstract & Executive Summary

Systematic quantitative strategies in liquid cryptocurrency markets frequently suffer from three structural pathologies: (1) curve-fitting to bull regimes via asymmetric long exposure, (2) catastrophic drawdown during prolonged choppy consolidations, and (3) hidden leverage bugs or synthetic proxy data contamination in backtesting engines.

This paper presents the formal technical specification, risk envelope, and empirical performance of JevTrader v5.3, an institutional trend-following strategy designed for 4-hour execution across an uncorrelated multi-asset cryptocurrency universe. JevTrader couples 14-period Donchian channel breakouts with a 200-period Exponential Moving Average (EMA) macro trend invariant, volume surge validation ($\ge 1.75\times$ 20 SMA), candlestick wick rejection gating ($\le 0.35$), and a dynamic 14-period Rate-of-Change (ROC) relative strength allocation tilt.

Capital compounding is powered by an asymmetric Convex Pyramiding Engine that initiates trades with a strict 2.0% equity risk budget, ratchets protective stops to breakeven (+0.1R) upon achieving +1.5R favorable excursion while adding up to 15% equity, and adds an additional 10% equity at +3.0R while locking in +1.5R profit. Portfolio drawdown is bounded by a 3.4 ATR trailing stop, a 32-hour stale breakout decay exit, hard leverage ceilings (2.50x basket, 1.00x single-asset), and an active rolling 20% peak-to-trough circuit breaker.

2. Market Inefficiencies & Strategic Philosophy

Cryptocurrency return distributions exhibit extreme leptokurtosis (fat tails) and acute volatility clustering. Under these mathematical distributions, trying to maximize predictive win rate is mathematically suboptimal. Instead, institutional alpha requires engineering extreme payoff asymmetry while capping left-tail losses.

• Low Win Rate, Massive Payoff: JevTrader does not attempt to be right 60% of the time. Over the 2-year regime, it achieved a 29.86% win rate, but paired it with a 5.00x Payoff Ratio (average win \$1,670.55 vs average loss \$334.15 on \$24.6k seed).
• Rapid Failure Recognition: False breakouts are pruned aggressively. If price fails to advance within 8 4-hour bars (32 hours) of entry, the position is liquidated at market, preventing capital erosion during choppy range-bound drift.
• Free-Roll Convexity: Trend expansions are heavily monetized via tiered additions that only execute once existing open risk has been completely eliminated by ratcheting protective stops into positive profit.

3. Mathematical Formalization & Execution Architecture

Let asset $k \in \mathcal{U} = \{\text{BTC}, \text{ETH}, \text{SOL}, \text{AVAX}, \text{DOGE}, \text{ADA}\}$ have continuous 4-hour OHLCV bar data at discrete time index $t$: $$B_k(t) = \left( O_k(t), H_k(t), L_k(t), C_k(t), V_k(t) \right)$$

3.1 Indicator Formulations

1. Macro Trend Invariant (200 EMA):

2. 14-Period Donchian Breakout Channels:

3. Volatility Normalization (14-Period ATR):

4. Volume Surge Validation:

5. Candlestick Wick Rejection Ratio:

3.2 Relative Strength Momentum Ranking

Every 4 hours, all assets without open positions are ranked by 14-period Rate of Change (ROC): $$\text{ROC}_k(t) = \frac{C_k(t) - C_k(t-14)}{C_k(t-14)}$$ The candidate universe is sorted in descending order. Sizing tilt $\omega_k$ is assigned: $$\omega_k = \begin{cases} 1.35, & \text{if Rank } 1 \\ 1.00, & \text{if Rank } 2 \\ 0.70, & \text{if Rank } \ge 3 \end{cases}$$

3.3 Position Sizing Formula

Given current portfolio equity $E(t)$, base risk fraction $r = 0.020$ (Long) or $r = 0.008$ (Short), stop loss distance $\Delta_{\text{SL}} = 3.4 \times \text{ATR}_k(t)$, and available basket capacity $K_{\text{avail}}(t)$: $$A_k(t) = \min \left( \frac{E(t) \times r \times \omega_k}{\Delta_{\text{SL}}} \times C_k(t), \; E(t) \times 0.75, \; K_{\text{avail}}(t) \right)$$ Where $K_{\text{avail}}(t) = \max \left( 0, \; 2.50 \times E(t) - \sum_{j \in \text{Active}} A_j(t) \right)$.

4. Convex Pyramiding Mechanics

Pyramiding in JevTrader is strictly gated by price excursion R-multiples, defined relative to initial stop distance $\Delta_{\text{SL}} = 3.4 \times \text{ATR}_k$:

• Tier 1 (Base Entry): Enters with $1.0R$ initial risk. Stop loss placed at $P_{\text{entry}} - 3.4 \times \text{ATR}$. Maximum allocation $0.75x$ equity.
• Tier 2 (+1.5R Excursion): Triggered when $C(t) \ge P_{\text{entry}} + 1.5 \times \Delta_{\text{SL}}$. $$\Delta A_{\text{Tier 2}} = \min(0.15 \times E(t), \; 0.90 \times E(t) - A)$$ $$\text{Stop Loss} \leftarrow \max(\text{Stop Loss}, \; P_{\text{entry}} + 0.10 \times \Delta_{\text{SL}})$$ Risk Impact: Convexity expansion with asymmetric base cushioning (protective ratchet stop placed at $P_{\text{entry}} + 0.10 \times \Delta_{\text{SL}}$).
• Tier 3 (+3.0R Excursion): Triggered when $C(t) \ge P_{\text{entry}} + 3.0 \times \Delta_{\text{SL}}$. $$\Delta A_{\text{Tier 3}} = \min(0.10 \times E(t), \; 1.00 \times E(t) - A)$$ $$\text{Stop Loss} \leftarrow \max(\text{Stop Loss}, \; P_{\text{entry}} + 1.50 \times \Delta_{\text{SL}})$$ Risk Impact: Total allocation reaches 1.00x equity while locking in +1.50R profit on the original capital base.

5. Adversarial Audit & Remediation Log

6. Multi-Asset Hexa-Core Expansion

Following audit remediation, candidate assets available on Kraken spot were audited over the 2-year regime:

Selection Rationale: While Quad-Core (BTC, ETH, SOL, AVAX) provided solid baseline trend-following, adding DOGE (retail momentum alpha) and ADA (clean breakdown shorts and low drawdown) expanded portfolio net profit by +71.0% while reducing maximum drawdown to 21.89% and lifting the Calmar Ratio from 2.80 to 4.55.

7. Exhaustive Empirical Results (2-Year Regime: 2024–2026)

8. Asset Contribution Matrix (Scenario A)

Pyramid Tier Attribution

• Tier 1 Base Entries: 255 trades (91.7% of trades) | 23.9% Win Rate | Net PnL: -$24,613.29 (Avg: -$96.52). Prunes false breakouts rapidly via 32h stale decay.
• Tier 2 Additions (+1.5R): 12 trades (4.3% of trades) | 91.7% Win Rate | Net PnL: +$19,575.55 (Avg: +$1,631.30). Free-roll exposure enlargement.
• Tier 3 Additions (+3.0R): 11 trades (4.0% of trades) | 100.0% Win Rate | Net PnL: +$78,534.49 (Avg: +$7,139.50). Generates >106% of total strategy net alpha by riding macro expansions.

9. Month-by-Month Performance Ledger (Scenario A)

10. Independent AI Verification Protocol (Private Audit)

Because JevTrader is a private, proprietary algorithmic trading system, external AI models (e.g. Claude, GPT-4o, Gemini Ultra) conduct independent audits via direct in-context audit dossiers or authorized local environment access.

# 1. Audit In-Context Equations & Invariants
- Base Risk Budget: Strictly capped at 2.0% equity per trade
- Tier 2 (+1.5R): Stop loss ratcheted to Breakeven + 0.1R (eliminates trade loss risk)
- Tier 3 (+3.0R): Stop loss ratcheted to +1.5R (locks in profit, expands allocation)
- Portfolio Basket Leverage: Hard ceiling at 2.50x equity

# 2. Local Deterministic Backtest Execution
python run_hexacore_2yr_full_metrics.py