↓ Skip to main content
Frank Zhang | AI & I

Investment

Welcome to the investment section. Here I share insights on quantitative trading strategies, Trade OS architecture, and financial planning.

The Core Design Philosophy of V7.2 Trade OS
#

The core design philosophy of V7.2 Trade OS can be summarized in one sentence:

Fully upgraded from a “rule-driven stock scanner” into a quantitative asset allocation and trading operating system powered by “AI Win Rate Scoring at its core, Control-Theory Phase Topology as its base, Signal Governance as its defense, and Single Source of Truth closed-loop micro-execution.”

The system completely eliminates common retail psychological traps such as “trading on intuition, relying on single indicators, chasing highs and panic selling, or turning large gains into drawdowns.” It is built upon Four Core Design Philosophies:

💡 Four Core Design Philosophies
#

1. Quantified Alpha & Objective Probability over Intuition
#

  • Statistical Win Rates Govern Expectation: The system abandons subjective bullish/bearish guesses. Via ai_winrate_engine.py, it converges Technical Trend (35%), Fundamental Consensus (30%), Micro Volume/Price (20%), and Executive Insider (15%) into a continuous 1-10 AI Win Rate and statistical outperformance probability over SPY/QQQ for 3 months (8-10 points maps to 70%-82% win probability).
  • Asymmetric Information Denoising: Through post_earnings_analyzer.py, it audits transcript sentiment and proprietary KPIs, while insider_radar.py tracks SEC Form 4 insider transactions to eliminate predictive biases with empirical data.

2. Control-Theory & Hysteresis over Noise
#

  • 2D Quantitative Topology: In acceleration_universe.json, it decouples 7 AI Value Chains from 4 Portfolio Defense Layers, eliminating traditional single-layer classification information collapse.
  • Control Theory Hysteresis & Denoising: In v6_acceleration_engine.py, it introduces a Bi-directional Schmitt Trigger Hysteresis Band and ULE (Unified Latent Energy) Manifold, eliminating 2nd-order derivative noise ($\Delta\Delta\text{Comp}$) so state transitions occur only during sustainable structural phase changes.

3. Signal Governance & Risk Sizing over Impulse
#

  • Context-Aware Capping: In ranking_engine.py, tickers in consolidation pullbacks or facing risk alerts receive forced score capping to eliminate low-quality breakout traps.
  • Smart Rebalancing Filter: The system overrides mechanical rebalancing—positions in structural breakdowns trigger REBALANCE BLOCKED (preventing catching falling knives), while positions in strong markup channels trigger REBALANCE HOLD (preventing premature profit taking).
  • Dynamic Budget Allocation: In exposure_engine.py, tactical (TRADE) budget is dynamically distributed by 6-factor signal scores, setting closed positions’ targets strictly to zero.

4. Single Source of Truth & Closed-Loop Protection
#

  • Eliminating Execution Contradictions: Operating through the Unified Action Terminal at templates/trade.html, backend endpoint /api/unified_recommendations (dashboard.py) arbitrates multi-engine signals into a single execution directive (BUY / ADD / TRIM / EXIT / HOLD / WATCH).
  • Profit Locking & Intraday Protection: In trailing_stop.py, it tracks Peak High after entry, activating a 5% dynamic trailing stop once floating profit reaches +10%; paired with breakout_detector.py intraday breakout monitoring, it resolves the profit roller-coaster dilemma.

🎯 Architectural Comparison Summary
#

DimensionV6.x Legacy DesignV7.2 Trade OS Core Philosophy
Decision BasisSimple technical lines & hardcoded rules1-10 AI Win Rate + Statistical Outperformance Probability
TaxonomyStatic single-layer classification2D Topology (7 Value Chains × 4 Portfolio Layers)
State DetectionCross-sectional rank vulnerable to noiseControl theory equations + Schmitt Trigger + ULE Manifold
Rebalance LogicMechanical buy-on-underweight regardless of trendSmart Rebalance (Breakdown Block / Strong Trend Hold)
Micro Risk ControlStatic fixed stop-loss5% Dynamic Trailing Stop + Intraday Breakout Detector
Action ConsoleDispersed views with occasional signal conflictsUnified Action Terminal (Single Source of Truth Arbitration)

Powered by Antigravity Vibe Coding:
From conceptual prototype to production deployment, this entire system was engineered through Antigravity Vibe Coding (AI Agent Pair Programming). Through high-frequency natural language interactions, complex trading philosophies were rapidly translated into rigorous code logic. The Antigravity Agent assisted in building the data pipelines, the signal fusion engine, and the defensive risk-management overlays, fundamentally reshaping the paradigm of quantitative infrastructure development.

2026