A regression betting system changes stake size after results; it does not change the odds or remove the house edge. The version examined here starts at 2 units, drops to 1 after the first win, then adds 1 unit after each further win. Any loss ends and resets the sequence.
Verdict In One Sentence
The system creates many short, easy-to-describe cycles and larger exposure during a winning run, but every roulette wager retains the same negative expected value; use the sequence only as a maths example, not as a route to profit.
Regression Is Not One Universal System
Casino writing uses “regression” loosely. It can mean reducing a large stake after a win, pressing a portion of profit, or following a named win-progression sequence. Two pages can therefore use the same label while describing different bets, reset rules, and targets.
A system cannot be tested until every rule is fixed. This guide uses one specific sequence found in older Casino Help material so the arithmetic can be checked. It should not be treated as the only definition used by players or casinos.
| Rule | Version tested on this page |
|---|---|
| Starting stake | 2 units |
| After the first win | Bet 1 unit |
| After each later win | Increase the next stake by 1 unit |
| After any loss | End the cycle and reset to 2 units |
| Bet type used for examples | European roulette red or black |
| Maximum sequence | Must be set before play; the uncapped sequence keeps growing |
How The 2-1-2-3 Sequence Works
- Bet 2 units on one even-money roulette option such as red.
- If it loses, the cycle finishes at minus 2 units. Reset or stop.
- If it wins, the cycle balance is plus 2 units and the next stake drops to 1 unit.
- After each additional win, increase the next stake by 1 unit: 1, then 2, then 3, then 4.
- The first loss ends the sequence. Calculate the actual result before deciding whether another cycle fits the limit.
“Even-money” describes the 1:1 payout, not a 50% chance. On a single-zero European wheel, red and black each cover 18 of 37 pockets. Zero loses for both bets, so the probability of winning one spin is 18/37, about 48.65%.
Worked Cycle Results In Units
| Results before reset | Stakes placed | Regression result | 1-unit flat result |
|---|---|---|---|
| L | 2 | -2 units | -1 unit |
| W, L | 2, 1 | +1 unit | 0 units |
| W, W, L | 2, 1, 2 | +1 unit | +1 unit |
| W, W, W, L | 2, 1, 2, 3 | +2 units | +2 units |
| W, W, W, W, L | 2, 1, 2, 3, 4 | +4 units | +3 units |
| W, W, W, W, W, L | 2, 1, 2, 3, 4, 5 | +7 units | +4 units |
The table includes the loss that ends the sequence. It compares the regression sequence with a constant 1-unit stake over the same results. The progression eventually produces a larger cycle win only because more money is being placed on later spins.
For a $5 unit, the first stake is $10. The five-win-then-loss row wagers $85 in total and finishes $35 ahead. Flat betting wagers $30 and finishes $20 ahead. The extra $15 result comes with $55 more turnover, not with improved roulette odds.
How Rare Is The Winning Run
On single-zero roulette, the chance of consecutive red wins is (18/37) raised to the number of spins. The calculation assumes a fair wheel and a fixed bet on red. Betting black gives the same probabilities.
| Consecutive wins | Probability of reaching that run | Approximate frequency |
|---|---|---|
| 2 | 23.67% | About 1 in 4.2 sequences |
| 3 | 11.51% | About 1 in 8.7 |
| 4 | 5.60% | About 1 in 17.9 |
| 5 | 2.72% | About 1 in 36.7 |
| 6 | 1.33% | About 1 in 75.4 |
| 8 | 0.31% | About 1 in 319 |
| 10 | 0.07% | About 1 in 1,346 |
These are probabilities, not a schedule. A six-win run is not guaranteed once every 75 attempts, and a long absence does not make the next run more likely. Each spin begins with the same 18 winning pockets out of 37.
The House Edge Applies To Turnover
A single-zero European roulette even-money bet has a house edge of 1/37, about 2.70%. The expected loss on a sequence is therefore about 2.70% of the total amount wagered, not 2.70% of the opening stake or deposit.
| Total wagered through cycles | Theoretical loss at 2.70% |
|---|---|
| 100 units | About 2.70 units |
| 500 units | About 13.51 units |
| 1,000 units | About 27.03 units |
| 5,000 units | About 135.14 units |
A progression can increase turnover without making that increase obvious. Later bets may feel like they use “profit,” but winnings become part of your balance as soon as they are credited. Risking them is still risking your money.
Expected Turnover Of One Uncapped Cycle
For the exact 2, 1, 2, 3 sequence on European roulette, the expected total stake before the first loss is about 3.845 units. Multiplying that turnover by the 1/37 house edge gives an expected cycle loss of about 0.104 units. With a $5 unit, that is roughly $0.52 per completed cycle in mathematical expectation.
The calculation does not predict the result of one cycle. Most cycles finish quickly, while a small number reach higher stakes. It combines all possible cycle lengths according to their probability. Adding a cap changes the distribution and expected turnover, but it does not reverse the sign of the expectation.
Why Short Cycles Can Feel Successful
The cycle often finishes after only one or two decisions. After a first-spin loss it loses 2 units. After one win followed by a loss it keeps 1 unit. That contrast can make the system feel as if it turns many sessions into small wins.
The feeling comes from the distribution of cycle results, not positive expectation. Losing cycles, larger later stakes, table limits, repeated resets, and continued turnover still matter. A system can change how wins and losses arrive without changing their long-run mathematical direction.
Regression Compared With Other Stake Plans
| Method | Stake movement | Main exposure |
|---|---|---|
| Flat betting | Same stake each round | Longer play keeps accumulating the house edge |
| 2-1-2-3 regression | Start high, reduce after first win, then add after wins | Opening loss is 2 units and long win runs create larger bets |
| Paroli or Reverse Martingale | Often doubles after wins and resets after a loss | Stake grows faster during a winning run |
| Martingale | Doubles after losses and resets after a win | Rapid growth during losing runs and collision with limits |
| D’Alembert | Adds after a loss and subtracts after a win | Slower growth can still build substantial exposure |
Positive progressions do not chase losses in the same mechanical way as Martingale, but “less explosive” is not the same as profitable or safe. They can encourage longer play because a winning sequence always appears one result away from a larger payoff.
A Cap Is Part Of The System
Without a pre-set cap, the stakes continue 2, 1, 2, 3, 4, 5, 6, 7 and upward while the player is winning. The probability of reaching high stakes declines, but the money at risk grows. Deciding the cap while excited defeats the point of having one.
- Set a maximum number of bets in one sequence.
- Set a maximum stake that includes chips needed for any allowed game action.
- Set a total session loss limit independent of the cycle result.
- Stop after the cap even if the most recent result was a win.
- Do not begin another cycle to recover a previous cycle.
- Do not raise the unit size because the table minimum feels too small.
Why Blackjack Is A Poor Fit
Blackjack hands do not all require the same final stake. Correct basic-strategy decisions can require a double down or split, so a progression unit can turn into two or more active wagers. Payouts can also differ for blackjack, surrender, insurance, and side bets.
Do not refuse a correct split or double solely because the progression stake has become uncomfortable. That means the unit or system does not fit the bankroll. Do not use insurance to protect a large progression bet; insurance is a separate wager with its own negative expectation under ordinary rules.
- Use the exact table rules and a matching basic-strategy chart.
- Budget for doubles and splits before choosing the base unit.
- Keep insurance and side bets outside the progression.
- Do not treat a blackjack win as evidence that the next hand is more likely to win.
- Stop if stake pressure changes a correct playing decision.
Common Claims Checked
| Claim | What the maths says |
|---|---|
| “It locks in profit after the first win” | A later loss can leave a cycle ahead, but repeated cycles still have negative expectation |
| “Larger bets use casino money” | Credited winnings are your money and create real exposure when re-staked |
| “A win streak can be ridden safely” | Later stakes grow, while the next result remains independent |
| “It survives cold streaks” | Each first-spin loss costs 2 units and repeated cycles accumulate losses |
| “It works better in blackjack” | Variable hand stakes and rule-dependent decisions make tracking more complex |
| “Stopping at a target beats the edge” | A target changes session length and result distribution, not expected value per dollar wagered |
How To Test The System Without Gambling
- Choose a fixed unit and write every rule, reset, cap, and stop condition.
- Use a spreadsheet or simulator with independent European roulette outcomes.
- Track total wagered, not only the opening bankroll and final balance.
- Run many samples and keep losing samples rather than showing one favourable path.
- Compare it with 1-unit flat betting over the same outcomes.
- Calculate expected loss as total turnover multiplied by 1/37.
- Repeat with American roulette at 2/38 to see the effect of the extra zero.
A simulation can show volatility, drawdown, stake peaks, and the spread of short-term outcomes. It cannot turn a negative-expectation wager positive. If code reports persistent profit, first check whether zero is handled as a loss, whether losses are omitted, or whether the sample is selected.
Safer Limits For Studying A Progression
The lowest-risk way to examine a regression system is with written examples or a no-money simulator. If real-money play is legal and chosen, the session needs limits that sit outside the progression so the sequence cannot rewrite them.
- Use only disposable entertainment money.
- Choose European single-zero roulette rather than double-zero roulette when comparing the same even-money bet.
- Keep the unit small enough that the maximum planned stake is comfortable.
- Set time, loss, and turnover limits before the first bet.
- Do not borrow, cancel withdrawals, or add funds to finish a sequence.
- Stop when a limit is reached even if the sequence is incomplete.
Last reviewed: 4 September 2026. This page analyses a defined staking sequence and does not recommend it as a way to make money.
Related Guides
- Roulette Rules
- Roulette Percentages And House Edge
- Anti-Martingale System
- Martingale System
- D’Alembert System
- Casino Bankroll Sizing Guide
Frequently Asked Questions
The term describes a stake plan that reduces or adjusts bets after certain results, but it has no universal sequence. This guide tests a 2, 1, 2, 3 progression that grows after consecutive wins and resets after a loss.
No. Changing stake size does not change pocket probabilities, payouts, or house edge. On European roulette, an even-money bet still has a 1/37 house edge on every spin.
Its stakes usually grow after wins rather than losses, so it avoids Martingale’s exact doubling pattern. It still has negative expectation and can create larger bets, extra turnover, and longer play.
The version analysed here ends the cycle after any loss and resets the next cycle to 2 units. Other systems using the regression label may use different rules.
No. On a fair roulette wheel, previous results do not make the next red or black result more likely. A win only changes the planned stake and current balance.
It is a poor fit because correct blackjack play can require doubles and splits, changing the money at risk. A progression should never cause a player to reject a correct basic-strategy decision or take insurance.
