This page compares three structurally different ways to manage your bets at the roulette table: the Martingale, the Fibonacci, and flat betting. Each one handles bet sizing differently, but none of them changes the underlying math of the game. Below, you’ll find out how each system works, what it asks of your bankroll, and how the three stack up in terms of escalation speed, table-limit sensitivity, and expected loss.
How Betting Systems Interact With the Mathematics of Roulette
A betting system tells you how much to stake on each spin. It does not influence where the ball lands. The house edge is a fixed property of the wheel, set by the ratio of losing outcomes to winning outcomes on any given bet type. No staking pattern, whether escalating, retreating, or constant, changes that ratio. You need to understand this before evaluating what any betting system can and can’t do for you.
The Fixed House Edge and Why Staking Patterns Cannot Change It
The house edge comes from the structure of the wheel itself. On a European wheel with 37 pockets and a single zero, the edge on an even-money bet is 2.7%. On an American wheel with 38 pockets and a double zero, it’s 5.26%. These numbers stay the same no matter how you size or sequence your bets. Changing your stake on spin five doesn’t change the probability of spin five producing a loss. What a staking system can do is redistribute outcomes across a session, concentrating wins and losses differently over time. That changes the variance and feel of play, but it doesn’t move the long-run expected loss in either direction.
Even-Money Bets as the Standard Context for System Comparison
All three systems are typically analyzed against even-money outside bets: red/black, odd/even, and high/low. These bets have the closest-to-even payout structure of anything on the table. That makes the progression math straightforward: each step in a sequence either wins one unit or loses one unit, so the main variable driving system behavior is the length and frequency of losing streaks, not payout multipliers.
Flat betting is the one system here that also gets discussed in the context of single-number wagers. A peer-reviewed simulation published through the UNLV Gaming Research Institute compared 100 Martingale rounds against an equivalent number of flat-stake single-number bets over 205 spins. The two strategies produced nearly identical expected losses, approximately -30.56 units for Martingale and -30.52 units for flat single-number betting, while their standard deviations diverged substantially, at 365 and 460 respectively. That result shows clearly that system choice governs variance, not expected value: the wheel’s math produced the same average loss under both approaches, while the spread of outcomes around that average was very different.
Martingale, Mechanics, Escalation, and Bankroll Pressure
Martingale is a negative-progression system built on one rule: double your stake after each loss, and go back to your base stake after any win. Bet growth is exponential, meaning each additional loss in a streak multiplies the required stake by two. That’s the fastest escalation of the three systems covered here. Even moderate losing streaks can push your stake to a number that looks nothing like what you started with.
The Doubling Rule and Its Recovery Logic
After any loss, the stake doubles. After any win, it resets to the base unit. The recovery logic is precise: a single win at any point in a losing streak recovers all prior losses in that streak and returns exactly one base unit of profit. That full-recovery-on-one-win property is what sets Martingale apart from Fibonacci, which steps back only two positions in its sequence after a win and therefore needs multiple wins to clear accumulated losses.
Martingale is often described as “guaranteed to recover,” but that guarantee depends on two things: enough bankroll depth, and a table maximum that can still accommodate the next required stake. Neither of those is guaranteed by probability alone. The recovery logic holds only as long as the doubling rule can still be applied.
Bet Escalation Under a Losing Streak
Under Martingale, bet size grows as 2ⁿ where n is the number of consecutive losses, making this the fastest escalation profile of the three systems. The figures below show how quickly required stakes and cumulative outlay grow during a losing streak. At a CA$5 base bet, surviving 8 consecutive losses requires a bankroll of approximately CA$1,275.
| Consecutive Losses | Required Stake (units) | Cumulative Bankroll Committed (units) |
|---|---|---|
| 8 | , | 255 |
| 10 | 1,024 | , |
Table Limit as a Structural Constraint
Martingale progression is bounded by the ratio of table maximum to table minimum, not just by your bankroll. To complete seven progression steps without interruption, the table maximum needs to be at least 128 times the table minimum. When a losing streak pushes the required stake above the table maximum, you can no longer apply the doubling rule, and all prior losses in that streak stay unrecovered. The table’s posted minimum and maximum set a hard ceiling on how many progression steps are actually available, regardless of how much money you’re holding.
Fibonacci, Mechanics, Sequence Progression, and Recovery Behavior
Fibonacci is a negative-progression system that moves stakes along the Fibonacci sequence after losses and steps back two positions in that sequence after wins. Compared to Martingale, stake growth is sub-exponential rather than exponential, and a single win produces only partial recovery of prior losses rather than full recovery. These two properties are connected: the slower escalation that makes Fibonacci less demanding on your bankroll is the same mechanic that prevents any one win from clearing the entire deficit built up during a losing run.
The Advance-One / Retreat-Two Rule
The rule has two parts: after each loss, move one step forward in the sequence; after each win, move two steps back. Because a win retreats two positions rather than resetting to the start, it cancels only part of the cumulative stake exposure built up during the preceding losses. This is the structural difference from Martingale, where a single win wipes the entire streak deficit and returns a one-unit profit. With Fibonacci, you trade Martingale’s faster recovery for a slower rate of stake escalation. The system reaches dangerous stake levels more gradually, but it also takes more wins to fully exit a losing sequence.
Escalation Speed and Cumulative Outlay Under a Losing Streak
Because Fibonacci grows sub-exponentially, the gap between its required stake and Martingale’s required stake widens substantially as a losing streak extends. The contrast is quantitative and drawn directly from the research.
| Consecutive Losses | Required Stake, Fibonacci (units) | Required Stake, Martingale (units) | Cumulative Outlay, Fibonacci (units) |
|---|---|---|---|
| 10 | 89 | 1,024 | 55 |
After ten consecutive losses, the next required Fibonacci stake is a fraction of the equivalent Martingale stake, and the total amount committed across the entire streak is lower still. That reflects the sequence’s slower climb, not any reduction in the underlying house edge.
Flat Betting, Mechanics, Variance Profile, and Long-Run Behaviour
Flat betting isn’t a progression system. It’s the deliberate absence of one. You fix your stake before the session starts and leave it unchanged regardless of whether the previous spin was a win or a loss. This produces the lowest variance of the three approaches covered here and the most predictable session outcome of any staking method you’re likely to encounter at a roulette table.
The Fixed-Stake Rule and Its Variance Consequences
Under flat betting, your stake never changes based on outcomes: no doubling, no sequence advancement, no adjustment of any kind. Because the bet never escalates, the variance of your cumulative result tracks directly with the underlying probability of your chosen bet type, without any amplification from a progression rule. The UNLV-published simulation by Pflaumer (2019) compared Martingale and constant-bet strategies over equivalent coups and found that the constant-bet approach on a simple chance produced a standard deviation of roughly 789 units against the Martingale’s 3,655 units over the same number of coups. That difference reflects escalation risk, not any change in expected loss. The practical upshot is that flat betting reveals the house edge more cleanly than either negative-progression system. No staking pattern is compressing wins into short sequences or inflating losses during streaks, so the underlying mathematical drag of the game shows up in your results without distortion.
Bankroll Predictability and Absence of Table-Limit Interaction
Because the stake never escalates, flat betting has no interaction with table maximum limits and no bankroll cliff triggered by consecutive losses. A losing streak of any length costs exactly one base unit per spin, the same as any other spin, so the rate of bankroll depletion is linear and determined solely by the house edge, not by streak length or sequence position. A flat bettor’s session length is a function of bankroll divided by expected loss per spin, and you can calculate that figure before the first spin. Neither Martingale nor Fibonacci offers this. Both introduce streak-dependent escalation that makes session length a function of variance rather than a predictable arithmetic outcome.
Side-by-Side Comparison of the Three Systems
The three systems differ across a fixed set of dimensions: progression type, escalation profile, bankroll pressure, table-limit dependency, recovery behavior on a single win, and impact on the underlying house edge. The differences aren’t marginal; they’re structural. No system alters the house edge. What varies is how each one distributes risk, bankroll demand, and recovery across a session.
The table below consolidates the mechanical and mathematical properties of each system across the dimensions covered in the sections above.
| Dimension | Martingale | Fibonacci | Flat Betting |
|---|---|---|---|
| Progression type | Negative | Negative | None (fixed stake) |
| Rule after a loss | Double the stake | Advance one step in sequence | No change |
| Rule after a win | Reset to base | Retreat two steps | No change |
| Bet growth profile | Exponential (2ⁿ) | Sub-exponential | Constant |
| Required stake after 10 consecutive losses | 1,024 units | 89 units | Base unit |
| Recovery on a single win | Full recovery of streak + 1 unit | Partial recovery | None (no recovery mechanic) |
| Bankroll pressure under losing streak | Highest | Moderate | Lowest |
| Table-limit sensitivity | High | Moderate | None |
| Effect on house edge | None | None | None |
| Variance profile | Highest | Moderate | Lowest |
Adjacent Progression Systems the Reader May Encounter
Martingale, Fibonacci, and flat betting are three points within a broader set of roulette staking approaches, not the full picture. When you research staking systems, you’ll regularly come across other named systems. These fall into two structural families: negative-progression variants, which escalate stakes after losses, and positive-progression variants, which escalate stakes after wins. This section covers the most commonly cited members of each family and the one interpretive point that applies to all of them.
Related Negative-Progression Variants
The D’Alembert system uses a gentler escalation rule than Martingale: add one unit to the stake after a loss, subtract one unit after a win. Over long sessions, this converges to the same loss rate as flat betting, producing lower short-session variance than Martingale but no long-run mathematical advantage. The Labouchère system is another negative-progression variant that shows up widely in staking system discussions. It operates through a cancellation sequence rather than a fixed multiplier or additive rule. Both systems trade off escalation speed and recovery behavior relative to Martingale and Fibonacci, but neither changes the underlying house edge. The same fixed expected-value constraint that governs Martingale and Fibonacci applies to every negative-progression variant.
Positive-Progression Alternatives
Positive-progression systems flip the escalation logic: stakes grow only during winning streaks, not losing ones. Named systems in this family include Paroli, 1-3-2-6, and Grand Martingale variants applied in reverse. Because losses occur at or near the base stake rather than at an escalated stake, downside exposure is structurally capped in a way that negative-progression systems don’t offer. The trade-off is that gains depend on winning runs actually materializing. When they don’t, the player cycles at the base stake without building up a recovery deficit. This reshapes the variance profile of a session, concentrating risk on missed winning streaks rather than on losing streaks, but it doesn’t change expected value. The house edge stays a fixed property of the wheel, regardless of which direction any staking system escalates.
Practical Bankroll and Table-Limit Considerations
Two practical questions follow from understanding how each system escalates stakes: how much bankroll you need before a session starts, and how often the streaks that exhaust that bankroll actually occur. The answers differ substantially across the three systems. Most players misread their exposure by looking at single-spin probability rather than how often damaging streaks recur across a full session. Evaluating bankroll requirements against realistic streak frequencies gives a more accurate picture of each system’s survivability under real playing conditions.
Streak Probability and Practical Ruin Risk
On an even-money bet at European roulette, the probability of ten consecutive losses is approximately 0.7%. That sounds low, but over 1,000 betting cycles a player will commonly encounter streaks of seven or more consecutive losses multiple times. A streak length that looks rare when viewed as a single probability becomes a near-certain feature of any extended session. Negative-progression systems need to be evaluated against how often damaging streaks recur across realistic session lengths, not against the probability of any individual losing spin.
Bankroll-to-Base-Bet Ratios by System
The bankroll multiple required to survive a realistic losing streak sets the practical floor below which each system breaks down. The benchmarks below, expressed as multiples of the base stake, show the operational threshold for each system.
- Martingale (highest-escalation system): A bankroll of at least 255 times the base bet is required to survive eight consecutive losses. At a CA$5 base bet, this is approximately CA$1,275.
- Fibonacci (sub-exponential negative-progression system): A cumulative outlay of 55 units is required to sustain a ten-loss streak, an order of magnitude lower than the highest-escalation system over the same streak length.
- Flat betting (fixed-stake system): Bankroll requirement is linear in session length and expected loss per spin, with no streak-driven escalation and no table-limit interaction.
