System Documentation · v1.0

Momentum Candle Backtester

A Streamlit-based momentum candle scanner + backtester for Binance altcoins. Three-step validation pipeline with purged walk-forward validation, adaptive ML, and dual-candidate AI verdict.

For Collaborators

Technical Architecture

// deep-dive · 8,264 lines of Python · reproducible

Full system internals. Every module, every constant, every design decision. Read this if you're going to touch the code.

  • Data → features → signals → validation pipeline
  • 72-method backtest · Purged WFO · Adaptive ML
  • Ratchet filter, regime weighting, NEUTRAL labeling
  • Why every guardrail exists (and what breaks if you remove it)
OPEN TECHNICAL DOCS →
For Traders

Trade Workflow

// how a signal becomes a trade · walkthrough

Follow one signal from scan to execution. What the app does for you, which gates it has to pass, and why you can trust the verdict.

  • Scanner finds it → 3-step validation → verdict
  • Candidate A vs B — the dual-method bake-off
  • Trade plan: entry zones, stops, TPs, management
  • Honest diagnostics: fill rate, OOS CI, regime fit
OPEN TRADE WORKFLOW →
Stack Streamlit · pandas · sklearn · Groq Universe 300+ USDT pairs (Binance) Methods 72 combinations / signal Validation Purged WFO + Bootstrap CI