SPO4Portfolio

Research sandbox for ETF portfolio backtesting with rolling retraining.

Structure

• Backtest.py: main rolling backtest entry.
• configs/: experiment settings.
• DataPipeline/: dataset and dataloader building.
• models/: prediction/allocation models.
• losses/: training objectives.
• portfolio/: portfolio optimization layer.
• optimizers/: optimizer factory.
• utils/: logging, metrics, plotting, tuning.

Quick start

1. Install:

   pip install -r requirements.txt

2. Check config (configs/spo_plus_linear.yaml or configs/softmax_linear.yaml).
3. Run:

   from Backtest import rolling_backtest
   rolling_backtest("configs/spo_plus_linear.yaml")

Outputs

Saved under outputs/<exp_name_timestamp>/:

• rolling_performance.csv
• rolling_weights.csv
• plots and logs