
I Built A Trading Bot With Claude That Made $168,236
- Goal: Test 12 famous trading strategies using Claude Fable (AI assistant) and find the winner.
- Approach: Convert strategies into Pine Script (TradingView) and apply to Bitcoin (USD) and NASDAQ.
- Result: Only one strategy outperformed buying & holding Bitcoin â the EMA crossover strategy. It generated $168,236 profit since July 2023.
- Define strategy â e.g., RSI Mean Reversion: rules for overbought/oversold, stop-loss, position sizing.
- Generate Pine Script â Claude writes the code (V6), including commissions and starting capital ($100,000).
- Fix errors â Screenshot the error, send it to Claude, get corrected version.
- Analyze results â Export CSV from TradingView, Claude explains win rate, max drawdown, and optimization options.
- Iterate â Adjust strategy (e.g., increase entry frequency, reduce risk).
| Strategy | Result | Key Insight |
|---|---|---|
| RSI Mean Reversion | Slightly negative (H4) | Optimized for daily â low signal frequency |
| EMA Crossover | +$168,236 (since July 2023) | Momentum strategy: long during uptrend, flat in downtrend |
| Buy & Hold Bitcoin | Underperformed | Only profitable in strong bull markets |
- Trend following: Only long positions in uptrends â no shorts.
- Exit logic: Trailing exit to lock in profits.
- Result: Since 2020 748% return (on $100,000 starting capital), profitable across all timeframes.
- Claude analyzes why a strategy works (e.g., âcut losses, let winners runâ).
- Improvement suggestions: Volatility-based position sizing, avoid over-optimizing entries.
- Warning: Simplicity wins â the baseline EMA version beat all âimprovedâ variants.
- Next step: Connect a trading MCP (e.g., Bybit) to Claude â provide API key, bot executes trades automatically.
- Workflow: Claude backtests thousands of strategies in parallel, flags profitable outliers.
- Key takeaway: AI-assisted backtesting is superior â without AI, you risk falling behind.
- Recommendation: Subscribe for the next video series on automated execution.
Note: No imperial units were present in the original text.






