Skip to content

QuantBT v1.2.0: Diagnostics-First WFO, Causal Meta and Reporting

Latest
Latest

Choose a tag to compare

View all tags
@BobbyAxerol BobbyAxerol released this 22 Oct 22:01
v1.2.0
d0e6bf8
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

QuantBT 1.2.0 / Native 0.4.3

A diagnostics-first WFO release with opt-in causal meta selection, shared samplers, researcher decision reports, explicit reporting populations, and account-correctness repairs.

Research And WFO

  • Shared legacy TPE, grouped multivariate TPE, CMA-ES and Sobol recipes, independent of whether meta selection is enabled. Mixed/conditional spaces require the documented explicit policies.
  • Mode 4 per-fold causal meta selection with matured-history boundaries, off/shadow/active controls, and route-specific accounting contracts. Nested Mode 1 is admitted on the documented scalar routes.
  • Opt-in Mode 5 calibration for one fixed parameter dictionary: sealed diagnostics/recipes, objective and selector families, winner qualification, regime/activity/probe/stress evidence, and explicit coordinate-only cross-mode assistance.
  • Shared pre-search diagnostics and fold/decision reports. No automatic sampler/selector router, universal superiority claim, or change to research defaults.

Reporting And Account Correctness

  • Optional Reporting V1 separates account-period, execution, closed-trade and closed-action populations with units, coverage, unavailable policies and drawdown episodes.
  • Corrected edge cases in financial clocks, funding coverage, contract multipliers, risk-reducing order admission, constraints, accepted positions, attribution and reporting horizons.
  • Unsupported non-default account/execution settings fail explicitly. Generic accounts remain gross-cross/netting; this is not isolated collateral, simultaneous hedge inventory or exchange-complete liquidation.

Installation

pip install --upgrade "quantbt-engine[optimization,viz,reports]==1.2.0"
# or
poetry add "quantbt-engine[optimization,viz,reports]==1.2.0"

On supported Linux x86_64/glibc, CPython 3.11-3.13, the exact pre-built quantbt-native==0.4.3 dependency installs automatically. No local Rust/Cargo compilation is required. The public native upload covers these three manylinux wheels, not Alpine/musl or every platform in candidate CI.

Compatibility And Evidence

  • Public endpoint factory names remain stable; documented optional WFO/reporting inputs are additive.
  • Account repairs can intentionally change results in affected edge cases. Read the ACC-01 migration guide.
  • Research evidence, software qualification, artifact certification and public installation proofs are distinct. Meta is not a promise of improvement for every alpha.
  • Performance remains workload-specific, not a universal speedup. See the README and benchmark methodology and full changelog.

Verified Publication

  • Published packages: quantbt-engine 1.2.0 and quantbt-native 0.4.3.
  • Native Release Certification: all four jobs passed on the immutable release tag.
  • Native PyPI publication: all three Linux x86_64 manylinux wheels published; their SHA256 hashes match the workflow artifacts.
  • Core PyPI publication: all five jobs passed; the public wheel and sdist hashes match the workflow artifacts.
  • Public Native Consumer Proof: 6/6 passed, Ubuntu 22.04/24.04 x CPython 3.11/3.12/3.13. Fresh Poetry projects resolved the published pair and exercised installed native execution, four sampler recipes, causal meta/W3 and exact sampler continuation. All six reports and their consumer log hashes were checked against the release verifier source.

On the supported Linux/glibc platform, normal installation resolves the pre-built native companion automatically; no separate Rust compilation is needed. This certifies software and public installation, not universal economic improvement or unsupported account/platform semantics.

Assets 2
Loading