stat_pairs_backtest/CHANGELOG.md

Changelog

All notable changes to this project are documented in this file.

Unreleased

No unreleased changes yet.

2026-07-30 v1.0.4

• Updated notebook and Panel analysis for the SP Quant result database schema, including explicit trading_instructions columns for action, assets, scaled disequilibrium, and beta.
• Changed selected-pair market charts to read from the market table and kept legacy packed instruction JSON support for older result databases.
• Added scaled_disequilibrium and beta to selected-pair theoretical execution displays.
• Improved VS Code notebook usability with the jupyter_bokeh dependency, direct Plotly figure rendering, and a dropdown Analyze control for individual pair selection.
• Made the Panel app use the dark theme by default and reduced the sidebar width from 430 px to 215 px with responsive sidebar controls.
• Expanded tests and notebook verification coverage for the new database schema and Panel layout defaults.

2026-07-29 v1.0.3

• Removed invalid fixed sizing mode from Panel Tabulator grids to avoid Bokeh layout warnings while preserving compact table layout.
• Changed the Panel Calculate action to refresh the result-file list before loading data and removed the standalone Panel Refresh button.

2026-07-29 v1.0.2

• Added a Panel application for single-day SPBT result analysis with result-file selection, minimum TARGET-change input, pair TheoRet table, pair selector, selected-pair execution table, and market/trade chart.
• Added a launcher script for the Panel application.
• Changed notebook and Panel pair analysis to use per-row Analyze actions from the Pair TheoRet grid, deferring selected-pair calculations until clicked.
• Adjusted Panel sizing so key controls use compact widths and Pair TheoRet uses content width with vertical scrolling instead of full-width paginated layout.
• Added a FastListTemplate shell to the Panel application for sidebar controls and configurable app color accents.
• Made Plotly chart panes use all available horizontal space.

2026-07-28 v1.0.1

• Added the spbt_day notebook for interactive single-day backtest result analysis, including SQLite result file selection from the local data directory.
• Added selector-pair loading and dense ranking by mr_score.final, preserving rows with invalid score JSON for inspection.
• Added theoretical return calculation for ranked pairs from trading_instructions, including reusable helper functions and tests.
• Added a Plotly histogram for visual analysis of total theoretical return by pair.
• Moved notebook support code into reusable scripts/spbt_day.py helpers.
• Adjusted notebook table outputs to show all relevant rows and reduce redundant intermediate displays.
• Added an alphabetically sorted pair selector for individual pair analysis.
• Added selected-pair theoretical execution tables and aligned TheoRet calculations with target-delta trade generation.
• Added per-asset strength values to selected-pair theoretical execution tables.
• Corrected theoretical execution size to use 10000 * strength / reference_price.
• Removed :USD quote suffixes from displayed pair names in notebook tables, chart hovers, and the pair selector dropdown while preserving full internal pair keys for calculations.
• Added num_trades to pair TheoRet summaries, counting asset-level theoretical trades from effective TARGET and CLOSE instructions.
• Added sortable interactive grids for the pair TheoRet and selected-pair theoretical execution tables.
• Styled interactive dataframe grids with black text on white backgrounds for readability across notebook themes.
• Added a selected-pair Plotly chart that overlays theoretical BUY/SELL executions on relative 1-minute market close data for both instruments.
• Anchored the selected-pair market chart at trading-day midnight and normalized relative prices to each instrument's close at that timestamp.
• Added a min_pctg_change threshold for ranked pair TheoRet calculations to skip small target-strength changes after a position is acquired.
• Added a notebook input field for the minimum TARGET strength-change threshold.

2026-07-25 v0.0.9

• Added contributing guidance and Python dependency declarations.
• Added placeholder files for active project directories.
• Updated ignore rules for local data, generated results, caches, and local environments.
• Documented unreleased changelog handling and release push behavior.

2026-07-25 v0.0.1

• Established the initial repository structure and project guidance.