How alfred-021/sp500_honest_backtest works
This repository is a reproducible S&P 500 classification experiment: load real or synthetic prices, derive past-only features and a five-day direction target, then compare shuffled and time-ordered validation against an always-up baseline. The main runner reports scores and saves results; a second runner compares several algorithms under the same protocol. A from-scratch NumPy neuron is included as an experimental classifier. The README also describes planned extensions, but they are not represented as implemented components. Edges reflect the shown calls and README; source outside the excerpts was not inspected.
Main entry points: compare_models.py Model comparison · run_backtest.py Backtest runner · evaluate.py Validation protocols · data.py Real data loader
Components 13
Experiment runners
- Backtest runner
run_backtest.py - Model comparison
compare_models.py
Market data
- Real data loader
src/data.py - Synthetic generator
src/data.py - Price cache
- Yahoo Finance
Features and models
- Features and target
src/features.py - Random forest
run_backtest.py - Comparison model suite
compare_models.py - NumPy neuron
src/neuron.py
Evaluation and output
- Validation protocols
src/evaluate.py - Score reports
results/results_sp500.csv
Other components
- User
Connections 20
- User → Backtest runner: runs
README.md - User → Model comparison: runs
README.md - Backtest runner → Real data loader: loads prices
run_backtest.py - Backtest runner → Synthetic generator: generates prices
run_backtest.py - Model comparison → Real data loader: loads prices
compare_models.py - Model comparison → Synthetic generator: generates prices
compare_models.py - Real data loader → Price cache: reads and writes
src/data.py - Real data loader → Yahoo Finance: downloads prices
src/data.py - Backtest runner → Features and target: builds inputs
run_backtest.py - Model comparison → Features and target: builds inputs
compare_models.py - Backtest runner → Random forest: configures
run_backtest.py - Model comparison → Comparison model suite: configures
compare_models.py - Model comparison → NumPy neuron: includes classifier
compare_models.py - Backtest runner → Validation protocols: evaluates
run_backtest.py - Model comparison → Validation protocols: evaluates
compare_models.py - Validation protocols → Random forest: fits and predicts
src/evaluate.py - Validation protocols → Comparison model suite: fits and predicts
src/evaluate.py - Validation protocols → NumPy neuron: fits and predicts
src/evaluate.py - Backtest runner → Score reports: writes scores
run_backtest.py - Model comparison → Score reports: writes scores
compare_models.py