How georgeqanfin-stack/gb-power-market-microstructure works
This repository is a quantitative research project for testing B6 boundary stress as a predictor of Scottish wind curtailment and GB Day-Ahead price direction, then evaluating a systematic trading strategy. The documented workflow collects NESO and Elexon data, explores it, classifies constraint regimes, derives directional signals, and evaluates evolving backtests. V4 adds macro filtering, rolling calibration, Wilson-score validation, and half-Kelly sizing. NIV analysis is a separate research branch. Source excerpts were unavailable, so links reflect README descriptions and notebook names; exact internal calls and data handoffs cannot be confirmed.
Main entry points: processed Processed data · 01_data_collection.ipynb Data collection · 05_backtest.ipynb Strategy backtest · 03_regime_classifier.ipynb B6 classifier
Components 13
Data research
- Data collection
notebooks/01_data_collection.ipynb - Processed data
data/processed - Exploratory analysis
notebooks/02_eda.ipynb
Signal research
- B6 classifier
notebooks/03_regime_classifier.ipynb - NIV analysis
notebooks/04_niv_analysis.ipynb - DA direction signal
notebooks/05_backtest.ipynb
Strategy evaluation
- Strategy backtest
notebooks/05_backtest.ipynb - Rolling calibration
notebooks/06_rolling_window_recalibration.ipynb - Macro filter
notebooks/07_macro_filter_backtest.ipynb
Other components
- Researcher
- NESO data
- Elexon data
- TTF prices
Connections 12
- Researcher → Data collection: initiates
- NESO data → Data collection: provides data
- Elexon data → Data collection: provides data
- Data collection → Processed data: writes data
- Processed data → Exploratory analysis: analyzed by
- Processed data → B6 classifier: trains on
- Processed data → NIV analysis: analyzed by
- B6 classifier → DA direction signal: provides probability
- DA direction signal → Strategy backtest: evaluated in
- Rolling calibration → Strategy backtest: recalibrates sizing
- TTF prices → Macro filter: provides prices
- Macro filter → Strategy backtest: filters signals