Sleeping Through the Volatility: Building an AI Agent for Automated Crypto Trading
summary
A fully automated crypto agent runs hourly price pulls via Binance API, stores data in SQL Express, and uses Python PyTorch Pandas to compute VWMA(20), MACD, RSI and candlestick patterns; signals weight trend 60% MACD 20% sentiment 10% candlesticks 10% and a trailing stop loss (typically 10%) enforces exits. Backtests show coin‑and period‑specific returns (Solana ~181% at 4h, ~300% at 8h, −24% at 12h). Development starts on laptop, moves to cloud (GitHub Neon, Render), claims ~15–25% average annual gain and currently pauses trading after the October 10 selloff.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 2:35Crypto trading automation need
- 8:35Humans vs machines trading
- 9:21AI agent for trading
- 10:02Building AI agent solo
- 13:16Eight steps of AI algorithm
- 16:29Secure crypto wallet setup
- 17:33Fetching prices and indicators
- 18:40Candlestick patterns explained
- 20:30Market sentiment analysis
- 23:09Trailing stop-loss protection
- 25:38Testing with TradingView
- 28:30Fine-tuning parameters
- 30:15Trend and MACD in algorithm
- 32:23Key takeaways
- 34:33Q&A
related talks
30mHow to build a stock market money printing machine with AI
33mPredicting Cryptocurrency Exchange Rates
23mComplex AI Forecasting Methods for Investments
48mPython's AI/ML for the Crypto-Industry
59mAI in Finance: Scope and Examples
48mScaling a cryptocurrency exchange using Data Science & Machine learning
50mAutomated algorithmic trading bot
07mQ n A with Dr Tom Starke