Upload metadata.json with huggingface_hub

metadata.json

@@ -28,23 +28,20 @@
     "romeo_keras_daily.keras"
   ],
   "metrics": {
-    "initial_capital": 500.0,
-    "final_capital": 884.8199412897085,
-    "cagr": 0.015843939373251237,
-    "annual_volatility": 0.5980092415124495,
-    "sharpe": 0.3257563389325535,
-    "max_drawdown": -0.482367849419699,
-    "total_trades": 3610,
-    "win_trades": 1786,
-    "win_rate": 0.49473684210526314,
-    "avg_pnl": 0.10659832168689985
+    "initial_capital": 100.0,
+    "final_capital": 162.31426062637883,
+    "total_return_pct": 62.314260626378825,
+    "total_trades": 207,
+    "win_trades": 102,
+    "win_rate": 0.4927536231884058,
+    "avg_pnl": 0.30103507548975383
   },
   "feature_list": "artifact['features']",
-  "usage": "Load artifact with joblib.load(). Align data to artifact['features'], fill missing with 0. Predict with ensemble weights. Use v6/backtest_v6.py with risk_per_trade=0.05 and threshold=0.4 for high-risk backtesting.",
+  "usage": "Load artifact with joblib.load(). Align data to artifact['features'], fill missing with 0. Predict with ensemble weights. Use v6/backtest_v6.py with risk_per_trade=0.25 and threshold=0.8 for high-risk backtesting. Tested under normal (commission 0.02%, slippage 0.5pips) and difficult conditions (commission 0.05%, slippage 1.0pips).",
   "training_data": "Yahoo Finance GC=F 15m intraday data with SMC and technical features",
   "evaluation_data": "Unseen fresh 15m intraday data",
   "frameworks": ["scikit-learn", "xgboost", "lightgbm", "tensorflow"],
   "python_version": "3.8+",
   "dependencies": ["joblib", "pandas", "numpy", "scipy"],
-  "caveats": "High-risk position sizing (5% risk per trade); higher volatility and drawdown; historical backtests only; not financial advice"
+  "caveats": "High-risk position sizing (25% risk per trade); higher volatility and drawdown; historical backtests only; not financial advice"
 }