Skip to content

v0.2.3 - Daily 213-Factor Validation Dashboard

Latest

Choose a tag to compare

@initial-d initial-d released this 29 Jul 08:54

v0.2.3 - Daily 213-Factor Validation Dashboard

This patch release promotes the daily AkShare CSI 300 full-factor validation into a first-class dashboard entry. It keeps the project validation-first: this is a public-data approximation with transaction costs, not an exact paper reproduction or investment claim.

Highlights

  • Added a README Validation Dashboard so the latest maintained public-data snapshot is visible from the project homepage.
  • Added docs/validation_dashboard.md as a stable dashboard entry point for the daily 213-factor validation result, interpretation, caveats, and acknowledgements.
  • Added the daily full-factor AkShare CSI 300 validation path in scripts/akshare_csi300_full_pipeline.py.
  • Updated package metadata to 0.2.3.

Validation Snapshot

The maintained public run uses current CSI 300 constituents, daily AkShare A-share data from 2021-01-04 to 2024-12-31, and the full 213-factor library.

At 7 bps effective cost:

strategy ann_return sharpe max_dd turnover cost_drag final_equity
equal_weight_daily 17.75% 0.882 25.76% 0.0010 0.13% 1.8744
factor_mean_daily 15.80% 0.701 32.51% 0.3627 49.16% 1.7579
factor_mean_buffered_daily 22.20% 0.919 27.76% 0.1397 18.94% 2.1616

The useful lesson is cost-aware: the naive daily factor signal has positive gross evidence, but excessive turnover consumes much of the edge. The buffered daily factor rule still evaluates scores daily while reducing unnecessary churn, allowing the factor portfolio to beat equal weight after costs in this public-data run.

Acknowledgements

Thanks to @redamancy231-create for contributing the AkShare zero-auth A-share data loader in PR #42. That data path made the CSI 300 public validation workflow possible without account registration. Thanks also to the AkShare project and maintainers for public Python access to A-share market data.

Caveats

  • This is not an exact paper reproduction; the paper data is not redistributed.
  • The run uses current CSI 300 membership resolved from public endpoints, not historical point-in-time membership.
  • The public AkShare path lacks industry, size, liquidity, beta, and production execution metadata.
  • Results are research diagnostics, not investment advice or live-trading evidence.

Verification

  • pytest tests/test_akshare_csi300_full_pipeline.py tests/test_akshare.py tests/test_public_data_validation.py
  • git diff --check