nullbench

Pre-register decisions. Score them against chance.
Lab / alpha — local seals under M1; not a global notary.

Product gate: without a green M1 gate, do not claim absolute「可稽核」or「永不 backfill」.
See docs/MATURITY.md.

nullbench is a null-first decision evaluation lab: freeze choices before outcomes, settle against equal-cost chance portfolios, keep an append-only ledger with semantic seals, and report descriptive percentiles plus sequential diagnostics.

It is not a lottery predictor. Negative expected-value domains are welcome as methodology stress tests.

Maturity (M0 → M4)

nullbench maturity
nullbench maturity --check-m1    # pytest -m m1

Install

pip install nullbench
# from source (needed for maturity --check-m1 tests)
pip install -e ".[dev]"

Python 3.11+.

60-second start

nullbench doctor
nullbench demo --name try1
nullbench report --study try1 --open
nullbench next --study try1
nullbench maturity --check-m1

Golden path

nullbench init my-study -d demo649
nullbench strategy add random --study my-study --tickets 5 --seed 1
nullbench freeze --study my-study --latest
nullbench settle --study my-study
nullbench report --study my-study --open

What M1 seals (local)

Residual risk: an adversary who rewrites all seals consistently still wins until M4. Documented in docs/THREAT_MODEL.md.

Commands

Docs

• MATURITY.md · INTEGRITY.md
• PRD.md · THREAT_MODEL.md · CLAIM_POLICY.md (M2 drafts)
• ARCHITECTURE.md · PRODUCT.md · CHANGELOG.md

Ethics

Pure simulation. No betting integration. Do not market predicted numbers.

License

MIT