A staking system determines how much you risk, not whether the underlying selection is good. Flat betting keeps the stake stable. Progressive systems change it after wins or losses according to a sequence. The difference can transform volatility and risk of ruin, but it cannot turn negative expected value into positive expected value.
This guide compares the main approaches using consistent examples, explains why loss-recovery systems fail under real constraints, and shows how to choose a method that keeps performance measurable.
Selection Strategy vs Staking Strategy
These are separate decisions.
- Selection strategy: estimates probability and decides whether the odds offer value.
- Staking strategy: decides how much to risk after a bet qualifies.
If a wager has estimated EV of −5%, changing the stake sequence does not make it +EV. It only changes when and how much you expect to lose.
For stake S, decimal odds O, and estimated probability p:
Expected profit = S × [(p × O) − 1]
The bracketed term is the expected return per unit. Stake size scales the result; it does not change its sign.
What Is Flat Betting?
Flat betting uses the same stake for every qualifying wager during a defined review period.
Example:
- bankroll: 1,000 units;
- flat stake: 10 units;
- every accepted bet: 10 units, regardless of recent results.
| Bet | Result | Stake | Profit at 2.00 | Bankroll |
|---|---|---|---|---|
| Start | 1,000 | |||
| 1 | Loss | 10 | −10 | 990 |
| 2 | Loss | 10 | −10 | 980 |
| 3 | Win | 10 | +10 | 990 |
| 4 | Loss | 10 | −10 | 980 |
| 5 | Win | 10 | +10 | 990 |
The stake is independent of the sequence. This makes the link between selection quality and results easier to audit.
Flat Currency Stake vs Flat Percentage
The term “flat betting” is sometimes used for two related approaches.
Fixed currency amount
Stake remains 10 units until a scheduled reset.
- simple records;
- stable sample comparison;
- relative risk rises after drawdown and falls after growth.
Fixed percentage of current bankroll
Stake is recalculated, such as 1% of the latest bankroll.
- exposure automatically contracts after losses;
- stake grows after gains;
- records require unit or percentage normalization.
Strictly speaking, the currency stake is not flat in the second method. Both avoid changing stakes because of a win/loss sequence, which is the key difference from progressive systems.
Advantages of Flat Betting
Transparent evaluation
When stakes are equal, profit and loss reflect selection outcomes more directly. A few oversized bets cannot dominate the record as easily.
Controlled exposure
The next stake does not explode after a losing sequence.
Less emotional feedback
The system gives no mathematical excuse to chase losses or press a winning streak.
Easier forecasting
Maximum planned daily, weekly and correlated exposure is simpler to estimate.
Compatible with testing
For a new model, flat paper stakes reveal hit rate, average odds, calibration and closing-line performance without variable sizing obscuring the evidence.
Flat betting still loses if selections are negative value. “Lower risk” is not “safe.”
What Is Progressive Staking?
A progressive system changes the next stake according to previous results.
Two broad categories exist:
- Negative progression: increase after losses, usually to recover them.
- Positive progression: increase after wins, attempting to exploit a streak.
Common examples include Martingale, Fibonacci, Labouchere, Oscar’s Grind, and Paroli. Their sequences differ, but none changes the probability or payout of the underlying bet.
Martingale
Classic Martingale doubles the stake after each loss and resets after a win, assuming even-money odds.
Starting at 10 units:
| Consecutive bet | Stake | Cumulative amount risked if it loses |
|---|---|---|
| 1 | 10 | 10 |
| 2 | 20 | 30 |
| 3 | 40 | 70 |
| 4 | 80 | 150 |
| 5 | 160 | 310 |
| 6 | 320 | 630 |
| 7 | 640 | 1,270 |
| 8 | 1,280 | 2,550 |
| 9 | 2,560 | 5,110 |
One win at true decimal 2.00 recovers previous losses and earns the original 10-unit target. The apparent reliability comes from exchanging many small wins for a rare catastrophic loss.
Real odds break the simple doubling rule
At decimal 1.91, a 20-unit win produces only 18.20 units of profit. After losing 10, that leaves +8.20, not the 10-unit target. To recover cumulative losses L and earn target T:
Next stake = (L + T) ÷ (Decimal odds − 1)
As odds shorten, required stakes grow even faster.
Why Martingale Fails in Practice
Finite bankroll
No bettor has unlimited capital. An eight-loss sequence can consume 2,550 units from a 10-unit starting stake before the ninth bet is placed.
Betting limits
Operators cap stakes and winnings. The required recovery bet may be rejected or exceed the market limit.
Odds are not always even money
Different prices require different multipliers. Simple doubling often fails to recover the target.
Bets are not independent
Several selections may depend on the same model error, team or market condition.
Execution risk
Prices move, events are voided, accounts are limited, and markets suspend.
Negative EV compounds
Increasing the stake after losses places the largest amount during the most dangerous part of the sequence without improving the next bet’s probability.
Fibonacci Progression
Fibonacci staking follows a sequence such as:
1, 1, 2, 3, 5, 8, 13, 21 ...
After a loss, the bettor moves forward; after a win, some versions move back two steps.
It grows more slowly than Martingale but retains the same central problem: stake size responds to past results rather than current edge. Recovery depends on odds, exact rules and eventually obtaining wins before the sequence becomes unaffordable.
Calling a progression slower does not make it mathematically profitable.
Labouchere and Cancellation Systems
Labouchere starts with a list of numbers. The next stake is typically the sum of the first and last values. A win removes those values; a loss adds the stake to the sequence.
The bookkeeping can create an impression of control, but the list does not influence event probability. During a bad run, both the sequence and stakes grow. Different user-defined starting lists also make risk hard to compare.
Complexity can hide exposure; it does not remove it.
Oscar’s Grind
Oscar’s Grind is a slower negative progression. Stakes usually increase by one unit after a win and remain unchanged after a loss until a one-unit cycle profit is achieved.
It avoids Martingale’s immediate doubling but can require long cycles and large cumulative turnover. At negative EV, additional turnover increases expected loss. A slow path to the same risk is still risk.
Positive Progression: Paroli
Paroli typically doubles after wins and resets after a loss. It aims to risk accumulated winnings rather than chase losses.
Example with 10 units at 2.00 and a three-win cap:
| Bet | Stake if previous bet won | Cycle result if outcome occurs |
|---|---|---|
| 1 | 10 | Loss: −10 |
| 2 | 20 | Win then loss: −10 overall |
| 3 | 40 | Three wins: +70 overall |
Positive progression caps loss more naturally than Martingale, but profits depend on consecutive wins. It still cannot change EV. It also increases stake after outcomes that do not make the next independent bet more likely.
Same Bets, Different Risk Profiles
Suppose ten fixed selections each have estimated probability 50% and odds 1.90. Their per-unit EV is:
(0.50 × 1.90) − 1 = −5%
| System | Expected effect | Main risk pattern |
|---|---|---|
| Flat 10 units | −5% of total stakes in expectation | Gradual variance |
| Martingale | Same negative EV per staked unit | Rare, very large loss |
| Fibonacci | Same negative EV per staked unit | Slower stake escalation |
| Paroli | Same negative EV per staked unit | Dependence on win streaks |
Because total stakes differ by realized sequence, final expected losses differ in amount. The underlying −5% return per staked unit remains.
The Gambler’s Fallacy
After six losses, many bettors feel a win is due. For independent events with unchanged probability, it is not.
If every selection has a 50% win probability:
P(Win next | six previous losses) = 50%
Past outcomes do not repair the next price. Progressive systems often formalize the gambler’s fallacy by making the largest bet after the longest losing run.
Sports bets may not be truly independent, but dependence can make matters worse if the losses expose a persistent model error.
Sequence Risk and Psychological Pressure
Progression changes not only mathematics but behavior.
As stakes grow, bettors may:
- skip required steps out of fear;
- select the next event hastily;
- accept a worse price;
- deposit additional money;
- alter the sequence after a partial recovery;
- hide losses or abandon records.
A system that can be followed only under ideal emotional conditions is not robust.
Flat Betting Is Not Completely Risk-Free
Flat stakes can still be too large. A fixed 10-unit stake is 1% of a 1,000-unit bankroll but 5% after the bankroll falls to 200.
Use scheduled review rules:
- reset monthly;
- reduce the unit after a defined drawdown;
- never increase during an initial model test;
- cap simultaneous and correlated exposure.
The stake should be small enough that a realistic losing run does not threaten essential finances or force a strategy change.
Variable Stakes Based on Edge
Not every non-flat method is a win/loss progression. Edge-based staking changes the amount according to estimated value rather than past results.
Examples:
- predefined confidence tiers;
- capped fractional Kelly;
- volatility-adjusted units.
These methods can be rational if probabilities are well calibrated, but they introduce model risk. Overestimating edge causes the largest stakes to be assigned to the most overconfident forecasts.
For an unproven model, flat small stakes provide cleaner evidence.
How to Compare Staking Systems Fairly
Backtests should use the same:
- chronological selections;
- accepted historical prices;
- starting bankroll;
- market limits;
- maximum stake rule;
- settlement and void handling;
- deposit prohibition;
- evaluation period.
Measure:
- return on total stakes;
- maximum drawdown;
- risk of ruin;
- largest individual stake;
- total turnover;
- exposure concentration;
- probability the system cannot place its next required stake.
Comparing only final profit hides the tail risk of progression.
Practical Decision Guide
| Situation | More defensible approach |
|---|---|
| New bettor learning settlement and records | Paper tracking or very small flat stake |
| New unverified model | Flat stake during out-of-sample test |
| Calibrated model with uncertain edge | Small fixed percentage or capped tiers |
| Demonstrated probabilities and strict controls | Fractional Kelly may be considered |
| Desire to recover recent losses | Stop rather than use progression |
| Stake sequence approaches bankroll or limit | End the sequence; do not deposit to continue |
Staking Checklist
- Does the method depend on current edge rather than previous result?
- Is the maximum possible next stake known?
- Can the bankroll survive a realistic losing sequence?
- Are operator limits included?
- Is correlated exposure capped?
- Does the system work at actual odds rather than ideal 2.00?
- Are deposits forbidden during a recovery cycle?
- Is performance measured by drawdown and turnover, not profit alone?
- Are pause rules defined before betting?
Frequently Asked Questions
Can Martingale guarantee a small profit?
Only under impossible assumptions: unlimited bankroll, unlimited stakes, exact required odds, flawless execution and eventual ability to place every bet. Real betting violates these assumptions.
Is Fibonacci safer than Martingale?
It grows more slowly, but it still increases exposure based on past losses and cannot change negative EV. “Slower” is not the same as safe.
Is flat betting always best?
It is transparent and suitable for testing, but no method is universally best. Flat stakes can still be oversized, and a validated edge-based approach may allocate capital more efficiently.
Should I increase stakes during a winning streak?
A streak does not improve the next independent bet. Increase only under a written bankroll or edge-based rule, not because recent outcomes feel favorable.
Does changing the stake affect the odds?
Usually not for small bets, but larger stakes can face lower limits, price movement, partial acceptance or account restrictions. Execution belongs in the risk model.
Final Thoughts
Flat betting keeps stakes independent of recent results and makes a strategy easier to evaluate. Progressive systems rearrange exposure around wins and losses, often replacing frequent small outcomes with rare large risks. None can create value where the selections have none.
Choose stake size after assessing price and probability, keep it small relative to bankroll, cap correlated exposure, and judge the method by drawdown and survival—not by a short profitable sequence. If the goal is to recover a loss, the correct next action is a pause, not a larger bet.
Last updated: July 2026
Published by LineScout Betting Academy
