Why no betting system can beat the house edge
A betting system changes the timing and size of wins and losses, never the house edge. In a ten million-round simulation, every progression finished negative once all wagers were counted.
The one-sentence answer
A betting system changes the shape of your results, not their average. They can make more sessions finish ahead, or reduce the chance of a dramatic swing, but they cannot remove the casino price built into each bet.
For definitions of the main progressions before reading the proof, start with the main guide to betting systems.
Our ten million-round simulation compared seven progressions plus a flat-betting control on the same single-zero even-money roulette bet. Every rule lost overall. The most exact match to the wheel price was D'Alembert, with a realised loss of 2.70% of everything wagered. Other systems landed above or below that in the sample, but none crossed into profit.
The rest of this page proves why that has to happen. It separates expected value from variance, shows the betting system maths, and then checks the theory against the full simulation table.
What a system can and cannot change
Most confusion comes from mixing up expected value, variance and session win rate. They are not the same thing.
Expected value is the long-run average result of a bet after weighting every possible outcome by its probability. For more worked examples, see worked expected value examples. Variance is the spread around that average, which is why two games with the same expected loss can feel very different. The volatility guide explains that spread in more detail.
A staking rule can make the ride feel smoother or sharper. It can decide whether losses arrive as many small dents or as a few hard hits. It can also alter the chance that a short session ends with a positive balance. None of that is the same as changing the value of the underlying bet.
Property | Can a system change it? | Why |
|---|---|---|
Expected value | No | The payout and probability set the price of the bet. A 2.70% roulette edge remains a 2.70% edge at every stake size. |
Variance | Yes | A progression can concentrate risk into fewer larger bets, or spread it across many smaller bets. |
Session win rate | Yes | A system can target frequent small winning sessions at the cost of rarer, larger losing sessions. |
Total amount wagered | Yes | Progressions often increase stakes after losses or wins, so the player may put more money through the same negative-price bet. |
Risk of ruin | Yes | Large jumps in stake can collide with the bankroll or table maximum before the recovery bet arrives. |
Betting system maths is therefore not about finding a hidden profitable pattern. It is about choosing which kind of volatility you prefer while accepting the same underlying price. A system that wins more sessions can still be worse for the bankroll if the losing sessions are much larger. A system with gentler losses can still lose if it keeps feeding money into a negative-expectation game.
The proof, in plain language
The formal statement
Linearity of expectation says the expected value of a sum is the sum of the expected values. This is the core gambling system proof.
Term | Meaning |
|---|---|
X_i | The dollar result of bet i. |
S_i | The stake chosen for bet i. |
e | The house edge per dollar, 2.70% for the single-zero even-money roulette bet used here. |
E | Expected value, the probability-weighted average result. |
Formula: E(total result) = E(X_1 + X_2 + ... + X_n) = E(X_1) + E(X_2) + ... + E(X_n). For each bet on the same negative-expectation game, E(X_i | past results) = -e × S_i. Add those terms together and the total stays negative whenever real money is staked.
The same thing without notation
Imagine every dollar placed on the wheel as buying the same losing ticket price. You may buy those tickets evenly, in a ladder, after losses, after wins, or in bursts. The price per dollar does not improve. If each dollar has negative expected value, a bag of those dollars also has negative expected value.
This does not mean every session loses. It means the average of many comparable sessions trends towards the game price. A player can beat the casino in a short session through luck. A staking rule cannot beat the casino mathematically.
Stopping after a win does not change the argument. It only chooses a flattering point to end the story. The losing expectation was paid on every wager made before the stop point. Restarting later begins another sequence of the same priced bets.
Why bet size is irrelevant
The same expected-value rate survives every stake pattern. The dollar loss changes only because the total amount wagered changes.
Staking pattern | Total staked | Expected value | Expected-value rate |
|---|---|---|---|
Flat $10 a spin for ten spins | $100 | -$2.70 | 2.70% loss rate |
Rising stakes, $2 up to $20 over ten spins | $110 | -$2.97 | 2.70% loss rate |
One $110 bet and nine $10 bets | $200 | -$5.41 | 2.70% loss rate |
The usual objection is that the system chooses the next stake after seeing previous results. That does not help. On a European wheel, the next spin is red with probability 48.65%, whatever came before. If the next stake is bigger after a loss, the player has only made the next negative-expectation term bigger.
The proof does not need the stake to be fixed in advance. It allows the stake to depend on what already happened. Once the next stake is chosen, that stake is still being placed into the same game with the same house edge.
The three false premises every system rests on
Outcomes are 'due'
The believer's claim is simple: after a long run of one colour, the opposite colour is due. That is the gambler's fallacy. Reports from Monte Carlo Casino on 18 August 1913 recorded 26 consecutive black results. A chosen colour appearing 26 times in a row on a European wheel is 1 in 136,823,184, yet the next red probability after any past sequence remains 48.65%.
Unlimited bankroll
The believer's claim is that the player can always outlast the losing streak. A $10 Martingale has $630 committed after 6 consecutive losses, and that streak has probability 1.834% on European even-money bets. That is not a remote fantasy. It is a normal feature of repeated play.
No table limits
The believer's claim is that the next double will always be accepted. With a $500 table maximum, the $10 Martingale is stopped after 6 losses because the next required stake is $640. The full ladder is shown in Martingale failure points.
Premise | Claim | Reality |
|---|---|---|
Outcomes are due | A run makes the opposite result more likely. | The next red probability on a European wheel is still 48.65%. |
Unlimited bankroll | Doubling can continue until recovery. | A $10 Martingale reaches a $630 net loss when the 6-loss streak hits. |
No table limits | The casino will always accept the recovery stake. | A $500 maximum blocks the $640 stake needed on bet 7. |
The law of large numbers is often misused in these arguments. It does not say a short-term imbalance must correct itself in the next spin. It says averages become more stable over many trials. For a negative-expectation game, that stability works against the player because the average settles around the house edge.
Simulation: 10 million rounds, 7 systems
Parameters
The simulation used the same wheel, session length, bankroll and table maximum for every staking rule. Seven systems were compared with a flat-betting control.
Parameter | Value |
|---|---|
Total rounds | 10 million |
Staking rules compared | 8 |
Rounds per staking rule | 1,250,000 |
Sessions per staking rule | 12,500 |
Spins in a session | 100 |
Base stake | $5 |
Starting bankroll | $1,000 |
Table maximum | $500 |
Wheel | Single-zero, even-money bet |
House edge measured against | 2.70% |
Results table
The master table shows the practical trade-off. High session win rates usually came with worse tail risk. Low drawdown systems still lost on average.
System | Sessions won | Mean result | Median result | Worst drawdown | Ruin rate | Total wagered | Actual loss |
|---|---|---|---|---|---|---|---|
Martingale | 69.8% | -$35.33 | $220.00 | -$1,000.00 | 28.81% | $19,880,735 | 2.22% |
Grand Martingale | 57.8% | -$47.27 | $350.00 | -$1,000.00 | 41.46% | $27,570,755 | 2.14% |
Fibonacci | 81.4% | -$30.16 | $75.00 | -$715.00 | 14.18% | $15,503,620 | 2.43% |
D'Alembert | 58.7% | -$79.23 | $65.00 | -$1,000.00 | 10.81% | $36,649,240 | 2.70% |
Labouchere | 55.3% | -$102.05 | $245.00 | -$1,000.00 | 40.86% | $50,267,180 | 2.54% |
Paroli, reverse Martingale, three steps | 39.6% | -$21.22 | -$25.00 | -$330.00 | 0.00% | $10,553,660 | 2.51% |
Oscar's Grind | 72.2% | -$40.52 | $55.00 | -$1,000.00 | 5.60% | $19,439,105 | 2.61% |
Flat betting, control | 35.8% | -$12.92 | -$10.00 | -$190.00 | 0.00% | $6,250,000 | 2.58% |
Read the mean and median together. A positive median can make a system feel safe because the middle session is ahead. Martingale and Grand Martingale both show that pattern. The mean is lower because the losing tail pulls the average down.
Flat betting looks less exciting because it does not manufacture as many positive sessions. It also produced the smallest worst drawdown in the table. That does not make it profitable. It only shows the difference between reducing stake escalation and removing the house edge.
What the drawdowns look like
The distribution table replaces the sales-page view with the risk view. Where the profit and loss shares do not fill the whole row, the remainder finished level.
System | Sessions ending in profit | Sessions ending in loss | Worst session result | Worst drawdown inside a session |
|---|---|---|---|---|
Martingale | 69.8% | 30.2% | -$1,000.00 | -$1,000.00 |
Grand Martingale | 57.8% | 42.2% | -$1,000.00 | -$1,000.00 |
Fibonacci | 81.4% | 18.3% | -$715.00 | -$715.00 |
D'Alembert | 58.7% | 41.2% | -$1,000.00 | -$1,000.00 |
Labouchere | 55.3% | 44.6% | -$1,000.00 | -$1,000.00 |
Paroli, reverse Martingale, three steps | 39.6% | 58.4% | -$325.00 | -$330.00 |
Oscar's Grind | 72.2% | 27.0% | -$1,000.00 | -$1,000.00 |
Flat betting, control | 35.8% | 56.2% | -$190.00 | -$190.00 |
This is the part most system pitches omit. The bankroll path matters as much as the final balance. A player who cannot tolerate a full-bankroll drawdown cannot sensibly judge a system only by its winning-session percentage.
Why the loss always lands near the house edge
One sentence covers it: expected loss = total wagered × house edge. The staking rule changes the total wagered. It does not change the price of the single-zero even-money bet.
System | Total wagered | Total loss | Actual loss as share of total wagered | House edge benchmark |
|---|---|---|---|---|
Martingale | $19,880,735 | $441,645 | 2.22% | 2.70% |
Grand Martingale | $27,570,755 | $590,905 | 2.14% | 2.70% |
Fibonacci | $15,503,620 | $377,060 | 2.43% | 2.70% |
D'Alembert | $36,649,240 | $990,360 | 2.70% | 2.70% |
Labouchere | $50,267,180 | $1,275,620 | 2.54% | 2.70% |
Paroli, reverse Martingale, three steps | $10,553,660 | $265,250 | 2.51% | 2.70% |
Oscar's Grind | $19,439,105 | $506,525 | 2.61% | 2.70% |
Flat betting, control | $6,250,000 | $161,460 | 2.58% | 2.70% |
Finite samples wobble. That is why some realised loss rates sit below the benchmark and some sit closer to it. The key point is that the result clusters around total wagered and the house edge, not around the name of the system.
This also explains why a more aggressive system can lose more dollars even when its percentage loss looks similar. More money is being pushed through the same priced game. The player may feel active and tactical, but the accounting line is still total wagering volume multiplied by the edge.
Why so many winning sessions get reported
System reports are usually session-level stories, not lifetime-level accounts. Fibonacci ended in profit in 81.4% of sessions and still averaged -$30.16 per session. Martingale ended in profit in 69.8% of sessions, with a median result of $220.00, yet its average session was -$35.33 because 28.81% of sessions ran out of bankroll.
That is a skewed distribution: many small wins, rare large losses. It also feeds a survivorship bias in what gets published. A short backtest can stop after a good run and look persuasive. A sales pitch can show sessions won, while hiding loss per dollar wagered. Betting system reviews that ignore total amount wagered miss the only denominator that matters.
The same effect appears in social posts and forum claims. A person is more likely to report a clean recovery sequence than a session that hit the table maximum. That selection bias does not change the maths, but it changes what readers see.
What actually moves the edge
A real edge change comes from a different game price, not a different staking pattern. None of the methods below are betting systems.
Method | What it changes | Effort required | How casinos respond |
|---|---|---|---|
Card counting | The edge itself, when the remaining cards favour the player | Accurate basic strategy, count conversion, bet variation and discipline. | Lower limits, faster shuffles or refusal of further blackjack play. |
Full-pay video poker | The pay table, which sets the return before a card is dealt | Perfect strategy and pay-table checking before play. | Pay-table changes, lower limits or game removal. |
Bonus EV | The value of the offer against the wagering it carries | Reading terms, clearing rules, game contribution and withdrawal conditions. | Offer limits, account reviews and promotion removal. |
Rebates and comps | Part of the theoretical loss comes back | High volume and careful tracking of theoretical loss. | Changed offers or reduced rewards. |
Skill-dependent blackjack choices | The cost of the hand: 0.48% house edge in our infinite-deck model, against several per cent for guesswork | Correct play for the exact rules. | Rule changes and table selection pressure. |
These methods try to change the price of the wager or select a better-priced wager. To compare legal low-price options without pretending a progression can help, use compare low-edge casino games.
The distinction is practical. A betting system says to alter stake size after wins or losses. Edge selection asks whether the bet itself is priced better before the money goes down. Those are different questions.
Systems sold online: how to evaluate the pitch
The audit that most pitches omit is simple. Ask what is being counted: sessions won, or money won per dollar wagered. Then look for these red flags.
- Short backtests that stop after a favourable sequence.
- No bankroll disclosure.
- Guaranteed language or claims that losses can always be recovered.
- A Martingale rebranded under a new name.
- Results shown per session instead of per unit wagered.
Example: a pitch says to start with $10, increase after every loss, stop after a small win, and ignore table limits. Broken down, it is not a new roulette system for sale. It is a loss-chasing progression with the same failure point: the next stake becomes too large for the bankroll or the $500 table maximum.
A fair pitch would show losing sessions, bankroll requirements, total amount wagered, and loss as a share of total wagered. If it does not, the headline win rate is not enough information to judge the risk.
What any system costs per hour
The same idea converts into time: rounds per hour × stake per round × house edge. The staking label does not appear in the formula.
Game | Rounds per hour | Average stake | Amount wagered in an hour | Expected hourly cost |
|---|---|---|---|---|
Single-zero roulette | 40 | $5.00 | $200 | $5.41 |
Double-zero roulette | 40 | $5.00 | $200 | $10.53 |
Online slots at 96% return | 400 | $0.50 | $200 | $8.00 |
Baccarat, banker bet | 60 | $10.00 | $600 | $6.36 |
For a gambling budget, this is the useful framing. Decide what an hour of play is allowed to cost before choosing a stake size. If a progression makes the average stake rise, the hourly cost rises with it.
The same stake can also cost different amounts across games. The cheaper choice is usually the lower house edge, not the more elaborate staking plan.
What this means for you
Choose the lowest-priced bet you can verify, size stakes from a fixed entertainment budget, and treat any system as a variance preference, not an edge. If the aim is to reduce mathematical cost, start with the house edge guide. True edge-changing play changes the game price. A progression does not.
Frequently asked questions
Can any betting system beat the house edge?
No. The one-line reason is that staking rules cannot change the expected value of individual negative-EV bets. If every dollar placed has a built-in house edge, changing the order or size of those dollars cannot turn the total expectation positive.
Why do betting systems win so often then?
Skewed outcome distributions: many small wins, rare large losses. In the simulation, Fibonacci ended in profit in 81.4% of sessions, but its average session result was still -$30.16. The wins are visible and frequent. The losses are less frequent, but large enough to dominate.
Is card counting a betting system?
No. It changes the edge itself rather than the staking pattern. A card counter is using information about the remaining cards, not hoping a progression beats a fixed negative price. See the advantage play note above for the distinction between edge-changing play and stake sizing.
Do betting systems work on slots?
No. Slots have no even-money bets, so most progressions cannot be applied at all. Changing the stake size does not alter the game return or the random result generation. A progression can change volatility and bankroll swings, but it cannot change a slot's expected value.