$32 Across Martingale: Rules, Probability, and Risk
$32 Across · Remove winning numbers · Three collections · Doubling risk
The $32 Across Martingale places minimum units across all six box numbers, removes each winning number, and ends after three different collections. If 7 arrives first, the original contributor proposed increasing the next across layout. The collection rule is understandable; the Martingale creates the danger.
Three collections occur before 7 about 44.40% of the time
Removing each winning number reduces the combinations still working. The probability is lower than the 51.2% chance of simply seeing any three box-number hits before 7 while all six bets stay up.
Starting $32 layout
| Number | Bet | Win | Action after win |
|---|---|---|---|
| 4 or 10 | $5 each | $9 each | Collect and remove that number |
| 5 or 9 | $5 each | $7 each | Collect and remove that number |
| 6 or 8 | $6 each | $7 each | Collect and remove that number |
- Confirm whether the $32 layout works on the come-out; the original explicitly said it should.
- When an active number hits, collect the win and remove that number’s Place bet.
- Ignore later rolls of a number already removed.
- After three different active numbers have paid, remove all remaining Place bets.
- Wait for the next 7 before beginning a new $32 cycle, if following the original reset rule.
What one completed cycle can produce
Because all original stakes are eventually removed after three collections, the cycle profit equals the three payouts: at least $21 if only 5/6/8/9 pay, and at most $25 if both 4 and 10 are among the winners ($9 + $9 + $7). A seven-out before the target loses only the Place bets still working; previously removed stakes and collected wins remain in the rack.
The Martingale escalation
| Level | Across amount | Total committed after full losses |
|---|---|---|
| 1 | $32 | $32 |
| 2 | $64 | $96 |
| 3 | $128 | $224 |
| 4 | $256 | $480 |
| 5 | $512 | $992 |
| 6 | $1,024 | $2,016 |
Real sequences often include one or two collections before 7, so the amount “needed to recover” is not always the full previous layout. That makes the original escalation ambiguous. Doubling mechanically still grows faster than most bankrolls and table limits.
Why simulation results are not proof
The old page cited one 4,800-roll software session that grew from $1,500 to more than $6,100. A single favorable simulation does not establish the long-run expectation. A useful test needs many independent runs, disclosed stop rules, exact commission/payout settings, and the number of bankrupt sessions—not only the best ending balance.
- Test the three-collection rule without progression first.
- Set a fixed maximum level rather than assuming unlimited doubling.
- Do not work the come-out unless that extra exposure is intentional.
- Track returned stakes separately from wins.
Compare the non-Martingale $22 Inside reduction and the general Place-bet guide.
Reproduce the 44.40% target probability
At the start, the six working numbers have 24 combinations against six combinations of 7, so the first collection arrives before 7 with probability 24/30, or 80%. After that hit, its combinations become neutral because the winning number is removed. The denominator for the next race therefore shrinks according to which number paid.
If 4 or 10 pays first, 21 active winning combinations remain and the chance of obtaining two more different collections before 7 is about 57.17%. If 5 or 9 pays first, 20 remain and that two-collection chance is about 55.72%. If 6 or 8 pays first, 19 remain and it is about 54.31%. Each figure also accounts for the second number being removed before the third race.
Weight those branches by their first-hit probabilities: (6/30 × 57.17%) + (8/30 × 55.72%) + (10/30 × 54.31%) = about 44.40%. Rolls of 2, 3, 11, 12, or an already removed number merely delay the decision. The result measures the three-distinct-hit exit rule; it does not include any later Martingale recovery attempt.
