Validation

As of March 13, 2026 for the published position, with repo organization updated on March 20, 2026.

BJW validation is organized into two independent tracks:

Track A - CVData parity: Does BJW produce the same output as CVData (Wattenberger)? Track B - Internal math and rules: Does BJW correctly implement published blackjack mathematics and basic strategy?

These are different questions and have different answers.

Current Status

Track	Status	Detail
Track A (CVData parity)	0/8 scenarios pass	March 13 rerun after two additional fixes still leaves residual gaps
Track B (Policy audit)	224/224 checks pass	All 16 play policy variants audited against published charts
Track B (Formula audit)	400/400 checks pass	Variance/risk outputs independently recomputed from raw accumulators
Track B (Theory matrix)	7/8 pass, 1 explained residual	Flat EV bands, directional invariants, monotonicity

What This Means

Track B passing means BJW correctly implements the math. Track A not passing means BJW and CVData still disagree on counted scenario outputs. We publish both results openly and treat Track B as the correctness claim.

Canonical Repo Sources

• docs/validation/README.md
• docs/validation/dossier/VALIDATION_STATUS.md
• docs/validation/cvdata/CVDATA_RECOMPARE_AFTER_FIXES_2026-03-13.md
• docs/validation/risk/OBSERVED_ROUND_VARIANCE_PRIMARY_REMEDIATION_2026-03-16.md

Navigation

-> Full validation position - complete disclosure including CVData delta table, root cause classification, and what we will and will not claim from BJW results

-> Audit detail - methodology and results for each Track B audit (policy, formula, theory)