Skip to content
vtp772002Public

About

Alpha search process – quantitative mathematical models on WorldQuant BRAIN

Topics

Resources

Stars

4 stars

Watchers

0 watching

Forks

Repository files navigation

AlphaFind

AlphaFind is an experimental Rust CLI for WorldQuant BRAIN research workflows. It can call BRAIN APIs, run simulations, screen an in-code candidate catalog, and calculate local portfolio diagnostics. It is not an official WorldQuant product, does not guarantee a passing submission, and does not reproduce the platform's private scoring algorithm.

Status and limitations

This is a research prototype. The submission gate is deliberately conservative: checks must all be present and explicitly passing before a local "pass" is reported, screening concurrency is bounded per account, Alpha and dataset reads paginate, and follow-up simulation URLs are restricted to the configured BRAIN host. Candidate formulas are loaded from an ignored local store rather than from tracked source.

Known gaps, stated plainly:

  • The local check evaluator is fail-closed (and submit includes a separate local portfolio-policy gate), but it is not version-aware and the expected core check list is hardcoded.
  • Simulation and submission POSTs are non-idempotent; ambiguous transport failures fail closed with a reconciliation fingerprint instead of being retried, and no automatic reconciliation exists yet.
  • Portfolio and PnL reads report portfolio_os.json and portfolio_pnl.json cache age, active/OOS classification counts, and support --refresh for live PnL re-fetching.
  • Git history still contains real field identifiers from earlier candidate catalogs. That is a release blocker for public distribution; see Architecture.
  • Research reports (radar, matrix, blueprint, shadow) use heuristics and local, user-owned PnL. They do not measure platform-wide co-movement or forecast official Uniqueness.

Current official BRAIN rules and verified submission criteria are summarized in BRAIN Rules and Submission. Confirm rules in the signed-in platform before submitting.

Features in this checkout

  • Rust CLI and asynchronous BRAIN API client.
  • Simulation, screening, parameter sweep, and submission commands.
  • Local PnL correlation, portfolio, impact, health, and HRP allocation/admission (allocate) diagnostics with --refresh controls.
  • Dataset census and heuristic research reports.

These are implementation features, not claims that results are complete, production-grade, or platform-approved.

Build from source

git clone https://github.com/vtp772002/alphafind.git
cd alphafind
cargo build --release
./target/release/alphafind --help

See Quickstart for credential handling and safe first steps. The installer (install.sh) checks for a clean working tree and updates via git pull --ff-only, setting 0600 permissions on .env.

Documentation

Privacy and account rules

Use one BRAIN account controlled by the participant, and do not share credentials or account IDs. Keep .env, candidate formulas, Alpha IDs, private PnL, and pending performance data out of version control and public logs.

Tracked source is free of candidate formulas: src/taxonomy.rs uses abstract placeholders and loads candidates from the ignored local store data/private_candidates.json. Downloaded field metadata (data/fields_*.json) is now untracked for the same reason.

The remaining privacy blocker is Git history: earlier commits contain real field identifiers from previous candidate catalogs. Deleting them in a new commit does not remove them, so rewrite history (or keep the repository private) before treating this checkout as safe to publish. scripts/sanitize_check.sh is a useful pre-push signal, not proof: it reports when it cannot run its candidate or Alpha-ID scans.

Authentication output may include account identifiers or email addresses. Sanitize logs before sharing them.

License

MIT. See LICENSE.

About

Alpha search process – quantitative mathematical models on WorldQuant BRAIN

Topics

Resources

Stars

4 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages