Executive Summary Platform

Architecture Map Interactive

About Greg

Silver Aperture Consulting Group

Blackjack Decision Lab

A rigorous blackjack learning application teaching probability, decision quality, and bankroll management — every claim traced to a source, every decision scored before its outcome.

Due Diligence Package · October 2026

0Tests

0Stages

0Glossary

0Views

Blackjack Decision Lab is not a casino game, a toy, or a gamification gimmick. It is a mathematically rigorous learning application backed by 605 automated tests at 100% pass rate, four independent benchmark fixtures verified against Wizard of Odds and BlackjackInfo, and a 46-lesson curriculum spanning 14 stages. Every strategy value traces to a source. Every decision is scored using information available at the time it was made — before the outcome is revealed. The goal is not to teach people to gamble. It is to teach decision quality under uncertainty.

Capabilities

What It Does

Six core capabilities, each engineered with the same rigor applied to the game engine itself.

♠

Play Table with Decision Scoring

Full blackjack table with real-time decision evaluation. Every action is compared to basic strategy before the outcome is known. Coached mode highlights the correct play.

📚

46-Lesson Curriculum

14 progressive stages from orientation through integrated proficiency. Each lesson includes plain explanation, worked examples, interactive demos, misconception correction, and drills.

🃏

Card Counting Training

Hi-Lo system with 4 practice modes: flash cards, hand counting, true count estimation, and full table simulation. 22 verified deviation indices.

📈

Monte Carlo Simulation

Run thousands of shoes to visualize bankroll trajectories, compare betting systems, and understand variance. Kelly sizing, flat betting, and progression analysis.

🔬

5 Interactive Labs

Bankroll Weather Map, Betting System Autopsy, Count Uncertainty, Rule Change Impact, and Other-Player Myth Buster. Each lab turns data into intuition.

🤖

AI-Powered Explanations

Claude adapter provides context-aware explanations within validated guardrails. AI explains — it never creates probabilities, indices, or bet sizes. Mathematical engines own the truth.

Architecture

Built for Rigor

A desktop-first application with offline capability, pure functional engines, and strict information boundaries.

Technology Stack

ShellTauri 2 + Rust

FrontendReact 19 + TypeScript

DatabaseSQLite (local)

SimulationRust Commands

Game Engine10 Pure Modules

StrategyCounting + Deviations

LearningCourse + SM-2 Review

EvidenceSources + Benchmarks

14 Engine Modules

Corecards · hand · rng · rules

Playsettlement · dealer · actions · round

Strategystrategy · shoe-manager · counting

Extendedcompanions · simulation · spaced-review

Evidence

Evidence & Rigor

Four benchmark fixtures verified against independent sources. 28 regression scenarios with pinned seeds. Every number is auditable.

370/370

BF-001: Basic Strategy

Every cell in the total-dependent basic strategy table verified against Wizard of Odds and BlackjackInfo. Zero discrepancies.

Verified

BF-002: Dealer Distributions

Dealer outcome probabilities integrated from Wizard of Odds Appendix 2b. Post-peek derivation verified for S17 rules.

-0.4026%

BF-003: House Edge

House edge verified against Wizard of Odds Appendix 9. Phase 1 preset: 6-deck, S17, DAS, late surrender.

22 Indices

BF-005: Hi-Lo Deviations

22 deviation indices verified against Wizard of Odds, BlackjackInfo, and first-principles derivation.

28 Scenarios

BF-004: Regression Suite

20 pinned-seed scenarios covering naturals, splits, doubles, surrenders, insurance, and edge cases. Plus 500-round soak test with card conservation.

14/14

Settlement Truth Table

Every settlement outcome verified: player natural, dealer natural, both natural, bust, push, surrender, insurance, split, double down, and all payout ratios.

Philosophy

Design Philosophy

Built on five principles that distinguish this from every casino app and blackjack trainer on the market.

Core Principles

⚖

Decisions Scored Before OutcomesEvery action is evaluated against basic strategy using only the information available when the decision was made. A correct decision that loses is recognized as correct. A mistake that wins is still a mistake.

🚫

No Slot-Machine TheatricsNo dopamine manipulation, no misleading win animations, no "hot streaks." XP and badges are restrained — adapted from enterprise LMS patterns, not mobile casino games.

📋

Every Claim Traces to a Source15 sources in the register, from Thorp's Beat the Dealer to Wizard of Odds. Conflicting sources create an issue for resolution — they are never silently reconciled.

🔒

Information Boundaries EnforcedThe player-facing API cannot access the hole card before reveal, the full shoe composition, or future draws. This boundary is tested and verified in the engine.

⏱

Calm CompetenceThe UI communicates confidence through clarity, not flash. Felt texture, clean card animations, keyboard controls. Professional-grade tools for serious learners.

Specifications

Technical Specifications

605

Automated Tests

Zero failures, 100% pass rate

46

Curriculum Lessons

14 stages, progressive mastery

~550

KB Bundle Size

142KB gzipped JS

64

Source Files

12 test files, strict TS

14

Engine Modules

Pure functional, deterministic

130

Glossary Entries

6 categories, cross-referenced

21

UI Views

Table, labs, course, drill

15

Verified Sources

Source register with status

"Good decisions can lose; bad decisions can win. We score the decision, not the outcome."

♠ Decision Lab Architecture Map

75 files 605 tests 46 lessons

Decision Lab 7 Domains

14 Engine Modules 21 Views 46 Lessons 140 Glossary 605 Tests

Developer Profile

Greg Collins

Full-Stack Engineer & Platform Architect

Silver Aperture Consulting Group, LLC

gregecollins@gmail.com (518) 669-3006

Professional Summary

Technologist, builder, and consultant with deep experience across full-stack engineering, enterprise platform architecture, and data-driven product development. Greg designs and builds production-grade systems that prioritize correctness, maintainability, and honest user experience.

His approach combines rigorous engineering discipline with pragmatic delivery — shipping real software that solves real problems, backed by automated tests and traceable evidence. The Blackjack Decision Lab exemplifies this philosophy: a mathematically verified game engine with 605 tests, a curated curriculum built on primary sources, and a user interface designed for calm competence rather than dopamine manipulation.

Technology Expertise

TypeScript

Pure functional engines, strict types

React 19

Component architecture, hooks, state

Rust / Tauri 2

Desktop shell, native performance

SQLite

Local-first data, offline capable

Claude AI

LLM integration with guardrails

Test-Driven Dev

605 tests, benchmark fixtures

Key Accomplishments on This Project

Designed and implemented a 10-module pure functional game engine with deterministic replay via seeded PRNG
Built 4 independent benchmark fixtures verified against Wizard of Odds and BlackjackInfo — 370/370 basic strategy cells, dealer distributions, house edge, and Hi-Lo deviation indices
Created a 46-lesson curriculum across 14 progressive stages with SM-2 spaced review adaptation
Engineered 5 interactive labs (Bankroll Weather Map, Betting Autopsy, Count Uncertainty, Rule Impact, Other-Player Myth Buster) that transform statistical concepts into visual intuition
Implemented strict information boundaries — the player-facing API provably cannot access the hole card, shoe composition, or future draws
Integrated Claude AI with prompt guardrails ensuring AI explains but never invents probabilities, indices, or bet sizes
Achieved 605 automated tests at 100% pass rate with 28 pinned-seed regression scenarios
Built a Monte Carlo simulation engine in Rust for bankroll trajectory analysis with Kelly sizing calculations

Contact

Email: gregecollins@gmail.com

Phone: (518) 669-3006