Research Ledger
Discoveries are published as auditable claims, not slogans. Each entry keeps the question, evidence, implication, and limitation visible so users can inspect the conditions behind the result.
Discoveries
Simulation Notes
D-01
Spread Size vs Variance
- Question
- How much does a wider bet spread increase EV relative to risk?
- Evidence
- Across representative shoe games, EV rises with spread width but variance and drawdown accelerate faster after moderate spread levels.
- Implication
- Risk-constrained users should optimize for EV per unit variance, not raw EV alone.
- Limitations
- Results depend on table limits and hourly rounds assumptions, which must match user context.
D-02
Penetration Sensitivity
- Question
- How strongly does deeper penetration influence edge?
- Evidence
- Deeper penetration materially increases playable high-TC rounds and improves long-run edge across most count systems.
- Implication
- Table selection can be more important than small strategy refinements.
- Limitations
- Observed gains assume stable rules and no countermeasures by casino staff.
D-03
True Count Bucketing
- Question
- What is the cost of rounded TC buckets versus exact TC handling?
- Evidence
- Coarser buckets simplify play but introduce measurable EV leakage around key deviation thresholds.
- Implication
- Training should target precision near threshold decisions before adding speed constraints.
- Limitations
- The magnitude of EV loss changes by ruleset and chosen strategy index set.
Definitions
Project Terms and Variables
Kelly Risk
A bankroll sizing model that scales bets from estimated player edge and variance. Fractional Kelly values, such as Kelly .25 or Kelly .75, deliberately bet below full Kelly to reduce volatility and drawdown exposure.
Expected Value (EV)
Expected value: the long-run average profit or loss for a decision, hand, round, hour, or betting ramp under the stated rules and assumptions.
Return to Player (RTP)
Return to player: total expected return divided by total wagered. An RTP above 100% means the simulated configuration has a positive player expectation before real-world constraints.
True Count
Running count normalized by decks remaining. This lets the same count signal be compared across different shoe depths and deck counts.
Penetration
The share of the shoe dealt before reshuffle. Deeper penetration exposes more late-shoe high-count rounds, which can materially change EV and variance.
Variance
A measure of outcome spread around the average result. Higher variance means results can swing farther from EV over realistic session lengths.
Drawdown
A decline from a bankroll peak to a later low. The planner treats drawdown as a practical risk measure alongside long-run expectation.
Bet Ramp
A schedule that maps count buckets to wager sizes. The ramp connects simulated edge estimates to bankroll, table limits, and bet increments.
Ruleset
The table rules used by the simulation, including H17 or S17, double after split, surrender, resplit aces, payout, deck count, and penetration.