EKG Sovereign — deep-RL trading engine (PPO)
Olaoluwa Bello
A PPO reinforcement-learning trading system built from scratch — data pipeline, custom Gymnasium environment, broker integration, risk engine.
Validated out-of-sample on locked holdout data: 63.6% win rate, +6.33% net on unseen candles. Deployed in live shadow mode — no hype, real audits.
Safety is deterministic: hard risk gates, a circuit breaker with cooldown, a daily-loss hard-stop and a kill-switch that survives restarts.
Code: https://github.com/KwabenaBellz/EKG-Empire — Stack: Python, Stable-Baselines3 (PPO), Gymnasium, pandas/NumPy, MT5 and Deriv APIs.
Like this project
Posted Jul 3, 2026
PPO deep-RL trading engine — validated out-of-sample (63.6% win rate on locked holdout), live shadow mode, hard risk gates and kill-switch.
Likes
0
Views
0
