Macro Factors vs. ML: Why We Chose Hard Logic Over Pattern-Fitting

[michaeljiangmingfeng-debug]

This is probably the most important question we get: "Why don't you use neural networks / deep learning / AI to find patterns?"

Short answer: because that way lies the black box.

Long answer: QuantToGo strategies are built on macro factor dynamics—real capital flows, behavioral incentives, regulatory boundaries. These factors are constant, reflecting objective reality.

The Core Difference

Machine Learning: Finds patterns in historical data. Patterns break when regimes shift. Your beautiful backtest becomes tomorrow's drawdown.

Macro Factors: Structural economic forces that don't rely on pattern-fitting. A limit-down rule will always trigger panic. Offshore RMB flows will always reflect global capital's real-time vote on China's economy.

Why This Matters

CNH-CHAU (offshore RMB long/short) works across decades—it's not a pattern; it's capital voting. The regulatory structure of China's capital controls creates two independent funding channels. This is structural. It doesn't change because an ML model found a new correlation.

How We Use AI

We use AI for auditing our data, not for mining patterns from historical noise. This is the courage of transparency: we expose our logic so you can verify it with ChatGPT, Claude, or DeepSeek.

We want you to audit us. That's the whole point.

The Trade-off

Yes, we sacrifice some potential alpha. An ML model might find short-term patterns we miss. But we gain:

• Explainability: Every position has a reason you can understand
• Persistence: Structural factors outlast pattern decay
• Auditability: You can verify our logic independently

🔗 Full strategy mechanisms: AI Hall
🔗 Try the strategies: Playground