Trade Vixion

The problem

Most trading tools show signals without ever proving whether those signals work after real trading costs.

What I built

• Watches 8 markets (gold, Bitcoin, Ethereum and 5 major forex pairs) across 9 timeframes with live prices
• Two independent forecasting engines run side by side and are compared head to head
• Every prediction is locked in a ledger before the outcome is known — nothing is edited afterwards
• Grades every prediction automatically and scores it in profit after trading costs, not just win rate
• Learns which model to trust on which market and timeframe, and refuses to signal where a model has a proven losing record
• "WAIT" is a normal answer: it doesn't force a trade
• Paper trading journal with position sizing; it never places real orders

How it's built

Node.js and MySQL backend with live market data feeds, an engine that analyzes the still-forming candle in real time, and a knowledge base of about 1,750 pattern rules.

• OANDA & Binance data
• Express
• MySQL
• React 19
• TanStack Router
• Google Gemini

My role · Designed, built, deployed and maintained solo with AI-assisted development (Claude Code).

Proof

• Live private app at tradevixion.site
• In replay testing on Bitcoin and Ethereum, it showed a modest positive edge after costs on 4-hour and daily charts — and no edge on 1–15 minute charts, which it reports honestly

Screenshots

Hiring for a leadership role? Book an interview