A fixed window of 6,000 consecutive rounds taken from the second simulation, since completed at 1,000,001 cards on each track — the equivalent of 100 hours at 60 seconds per completed round. Two OddsOfBlackjack bet-sizing rules are each compared with Basic Strategy — Flat Bet over the same number of rounds.
Retrospective analysis of simulated data — not real play and not a new simulation. Bets use a fixed $10 unit, not the balance-based sizing of the current app. The OOB and Basic Strategy tracks follow separate card paths and do not play identical hands. One 100-hour window is a single path, not a forecast or guarantee.
R is the realised net result inside the window: everything won minus everything staked. Every curve starts at $0 at hour 0. Both charts use the same scale, and Basic Strategy — Flat Bet is the same curve in each.
| Measure | OOB Graduated | OOB Selective | Basic Strategy — Flat Bet |
|---|---|---|---|
| Realised net result R (USD) | +6,745 | +4,050 | +2,450 |
| Difference versus Basic Strategy — Flat Bet (USD) | +4,295 | +1,600 | reference |
| Model contribution C (USD) | +77.80 | -22.72 | not available |
| R − C (USD) | +6,667.20 | +4,072.72 | not available |
| Maximum drawdown of R (USD) | 990 | 740 | 525 |
| Lowest R in the window (USD) | -30 | -70 | -225 |
| Highest R in the window (USD) | +6,775 | +4,080 | +2,490 |
| Initial stakes (USD) | 74,670 | 58,260 | 60,000 |
| Total staked, including doubles and splits (USD) | 81,950 | 63,810 | 65,740 |
| Largest initial bet (USD) | 120 | 130 | 10 |
| Rounds with a bet / observed without a bet | 4,358 / 1,642 | 4,358 / 1,642 | 6,000 / 0 |
| Deposits in the window | 0 | 0 | 0 |
| Balance at the start → end of the window (USD) | 3,855 → 10,600 | 3,135 → 7,185 | 2,225 → 4,675 |
| History before the window: net result / deposits | +1,855 / 1 | +135 / 2 | +225 / 1 |
Each account enters the window with its own history in the simulation: the starting balance is reported as it was, with no silent reset to $1,000. R in the charts and in the table is only the result inside the window. C is available for the OOB accounts only; for Basic Strategy no EV model is attributed, so C is not available. R − C is the realised result minus the model contribution; it is not attributed entirely to the variation of the cards.
The window was fixed in a manifest before the window-specific results were calculated, and it was not adjusted afterwards. The manifest was built only from round numbers and card positions. The simulation was already partly reviewed at that time, so this is not a blind selection on unknown data.
All stakes use a constant $10 unit and do not depend on the balance. The current app sizes its unit from the balance you confirm; that version is not tested here.
Funding: $1,000 per account at the start of the simulation; $1,000 is added only when a stake cannot be covered. No deposit was needed inside this window.
| Dataset | Second simulation, completed with 1,000,001 cards on each track (the window was fixed while it was running and uses completed batches only), seed 20261006, two tracks with the same shuffles: OOB and Basic Strategy |
| Game rules | 8 decks, cut card at 208 of 416 (50% penetration), dealer stands on soft 17, no hole-card peek, double on any two cards, no double after split, up to 2 hands from a split, no resplitting aces, no hitting split aces, no surrender, blackjack pays 3:2 |
| Engine build | OOB-ENG-b0681ede84a3 |
| Simulator | OOB-M6-de23a88e2d14 |
| Identification | OOB-M49-REPLAY100-v1 |
| Window manifest (SHA-256) | 2c50f5b13fdba8d0814ce5b68268fc37851d0e21eaf8a9f6d5835e3febe377ae |
| Accounting check | For every account, R equals the change in net result and the balance change equals R plus deposits. |