Production-Ready Spot Trading Bot

Develop a production-ready spot trading bot for Binance using Freqtrade and FreqAI. The bot should utilize a machine learning model, such as LightGBM or CatBoost, to predict coin prices. The project requires expertise in AI and trading systems to ensure efficient and reliable operation. Core specs: Framework: Freqtrade (latest) + FreqAI enabled Exchange: Binance Spot only ML Models: LightGBM or CatBoost Main timeframe: 5m / 15m Universe: Top 150 coins by market cap on Binance Max concurrent positions: 1 (up to 2 if low correlation) Position size: $500–$2,000 USD, scaled to model confidence Entry: model predicts ≥5% gain in 36h + confidence ≥78% + liquidity + no high-risk regime signals Stop-loss: hard -1% from entry Time exit: automatic close at 36h max holding period Dynamic exit: combo of RSI divergence, MACD, volume, EMA cross, Supertrend, price action (2–3 confirmations required) ML / Training requirements: Training data: 12–18 months of OHLCV + order book imbalance + BTC/ETH dominance + multi-TF indicators Retraining: every 5 days; hyperparameter optimization every 4–6 weeks Walk-forward validation; champion/challenger model approach Performance degradation alerts Infrastructure: PostgreSQL + Redis Docker setup Telegram/Discord alerts with prediction data Real-time performance dashboard Dry-run: minimum 2–3 weeks before live Deliverables I expect: PredictiveSwing36h.py – complete Freqtrade strategy file Optimized FreqAI config Backtest results (Win Rate, Profit Factor, Max Drawdown, Sharpe, Avg Trade Duration) on ≥12 months of data Docker deployment setup Documentation for retraining and hyperparameter tuning