Bankroll and Variance: Why a Good Bettor Can Lose for Several Weeks

In sports betting, one of the most common mistakes is assuming that a good bettor should win consistently.

A few losing days can feel worrying.

Two negative weeks can make you question a strategy.

A difficult month can even create the impression that an Edge has completely disappeared.

Yet that conclusion can be entirely wrong.

A bettor with a genuine mathematical advantage can still lose for several days, several weeks, or across dozens or even hundreds of bets.

The reason is a fundamental concept:

variance.

A profitable long-term strategy does not generate returns in a straight line.

Actual results fluctuate around their expected mathematical performance.

That is precisely why Bankroll Management is essential.

In this series, we continue to use our reference example:

Starting bankroll: €5,000

1U = €50

which means:

1% of the starting bankroll

In this article, we will explain why a good bettor can go through long losing periods, why a few weeks of results are not enough to judge a strategy, and how a properly sized bankroll helps survive variance.

What is variance in sports betting?

Variance describes the natural fluctuations of actual results around expected performance.

Suppose a bettor is able to identify bets with genuinely positive Expected Value.

That means the process should generate positive returns over the long term.

But it does not mean every bet will win.

It also does not mean every week or every month will be profitable.

A strategy can have positive expectation and still experience:

• several consecutive losses;
• a strongly negative week;
• several weeks without progress;
• a significant Drawdown;
• a long period around break-even.

That is simply how probabilistic environments work.

Positive Expected Value does not mean immediate positive results

Let us take a very simple example.

Suppose a bet has:

a true 55% probability of winning

at odds of:

2.00

The Expected Value is positive.

Over the long run, this type of bet is theoretically profitable.

But a 55% win probability also means:

a 45% probability of losing each bet.

An individual bet can therefore easily lose.

Several bets in a row can also lose.

The Edge exists across a large series of bets.

It does not guarantee the result of the next bet.

A good bet can lose

This is one of the most important concepts to understand.

A bet can be:

mathematically good

and still be:

a loser.

The opposite is also true.

A bad bet can win.

The outcome of one individual bet therefore does not determine whether the decision was good.

Suppose an event has a true probability of:

60%

and a bookmaker offers odds of:

1.80

The approximate implied probability of those odds is:

55.56%

If our 60% estimate is correct, the bet theoretically offers Value.

But there is still:

a 40% chance that the bet loses.

A losing result does not automatically mean the decision was poor.

Outcome and decision quality are two different things

In sports betting, it is essential to separate:

decision quality

from:

decision outcome.

A good decision can produce a loss.

A poor decision can produce a profit.

In the short term, randomness can have a major influence on results.

Over the long term, if the Edge is real, the quality of decisions should gradually become visible in the results.

That is exactly why a strategy should be judged over a sufficiently large number of bets.

Why a few bets mean almost nothing

Imagine a strategy with a true long-term ROI of:

+5%

That does not mean it will show exactly +5% after 20 bets.

It might show:

+18%

or:

-22%

or:

+2%

The dispersion can be very large.

With only a few dozen bets, statistical noise can be much larger than the Edge itself.

The smaller the sample, the more randomness can dominate the observed result.

Example with a €5,000 bankroll

Let us return to our reference:

Bankroll: €5,000

1U = €50

Suppose we use Flat Betting.

Each bet represents:

1U

Now imagine this sequence:

Result after 10 bets:

-4.15U

With:

1U = €50

that represents:

-€207.50

The bankroll falls from:

€5,000 to €4,792.50

That tells us almost nothing about whether the strategy is good or bad.

Ten bets are an extremely small sample.

Is a losing streak abnormal?

Not necessarily.

Even when a bettor wins more often than they lose, losing streaks are inevitable.

Suppose a strategy has:

a 55% win rate

That also means:

45% of bets lose

Across several hundred or several thousand bets, it is perfectly normal to observe sequences such as:

3 consecutive losses

5 consecutive losses

or more.

The more bets you place, the more likely you are to encounter at least one long negative sequence.

This can seem paradoxical.

A strategy can be excellent over 5,000 bets while still containing several very difficult periods.

The more you bet, the more bad runs you will encounter

Consider two records.

Record A

50 bets

Record B

5,000 bets

The second record is far more likely to contain:

• 5 consecutive losses;
• 8 consecutive losses;
• several Drawdowns;
• several negative weeks.

That does not mean Strategy B is worse.

It simply means it has been exposed to variance for much longer.

Any strategy that exists long enough will almost inevitably go through statistically uncomfortable periods.

Why one negative month proves very little

For many strategies, one month represents relatively few bets.

Take a tipster placing:

100 bets per month

A negative month therefore represents only:

100 observations

For a strategy with a relatively small Edge, that may be far too few to clearly separate:

skill

from:

variance.

One month at:

-8U

might be followed by a month at:

+15U

and then another at:

+4U

The cumulative result after three months would then be:

+11U

The first month was not necessarily evidence that something had gone wrong.

It may simply have been part of the normal distribution of results.

Can a strategy lose for several months?

Yes.

That depends on factors such as:

• its true Edge;
• its average odds;
• its win rate;
• its variance;
• its betting volume;
• the dispersion of results;
• its Stake size.

A strategy with a small but positive expected ROI can go through very long periods without meaningful progress.

For example, an Edge of:

+2%

can be valuable across several thousand bets.

But over 50 or 100 bets, it can be almost invisible.

Short-term variance can easily overwhelm that advantage.

The smaller the Edge, the harder it is to observe quickly

Suppose we have two strategies.

Strategy A

Expected true ROI:

+10%

Strategy B

Expected true ROI:

+2%

With comparable variance, Strategy A's advantage should be easier to detect statistically.

For Strategy B, the signal is much weaker.

A much larger sample will usually be needed to separate:

the signal

from:

statistical noise.

This is a fundamental concept in quantitative sports betting analysis.

Variance also depends on the odds

Not all strategies have the same variance profile.

A strategy mainly betting around:

1.50

will not behave exactly like one regularly betting:

4.00

or:

6.00

Higher-odds bets generally have a lower win rate.

That means winless periods can be much longer.

Let us take a simplified example.

Strategy A

Average odds:

1.60

Relatively high win rate.

Strategy B

Average odds:

4.00

Much lower win rate.

Even if both strategies have positive Expected Value, Strategy B can experience much longer losing streaks.

The bankroll must therefore match the strategy's risk profile.

Example: a high-odds strategy

Suppose a strategy wins:

28% of its bets

with average odds of:

4.00

The probability of losing each bet is therefore:

72%

Long losing streaks are entirely possible.

A sequence such as:

Loss

Loss

Loss

Loss

Loss

Win

is not unusual.

A winning bet at 4.00 produces:

+3U

profit on a 1U stake.

But the path of returns will naturally be much more irregular than for a low-odds strategy.

Win rate alone does not measure strategy quality

A bettor who wins:

70% of bets

is not necessarily better than someone who wins:

40%

It depends on the odds.

A 70% win rate at average odds of 1.20 may be insufficient.

A 40% win rate at average odds of 3.00 may be very profitable.

Variance must therefore be analysed alongside:

win rate AND odds.

What is Drawdown?

A Drawdown is the decline between a bankroll peak and the lowest point reached afterwards.

Suppose a strategy moves from:

+25U

down to:

+10U

The Drawdown is:

15U

With our reference:

1U = €50

that represents:

€750

or:

15% of the starting bankroll

A Drawdown does not automatically mean the strategy has become bad.

It may simply reflect a normal unfavourable period.

Example of a profitable strategy over time

Imagine the following progression:

At the end of Month 2, the strategy is at:

+14U

It then falls to:

+3U

The Drawdown is therefore:

11U

At that point, some users might believe the strategy has stopped working.

Yet three months later, it reaches:

+18U

The Drawdown was simply part of the path.

A bankroll curve is never a straight line

This is one of the most important mental models to abandon.

A profitable strategy does not normally look like:

↗ ↗ ↗ ↗ ↗ ↗ ↗

It looks more like a succession of:

rises

declines

flat periods

new highs

Drawdowns

and sometimes:

new highs again

Profitability must be assessed across the entire trajectory.

Not one week at a time.

Why the bankroll must absorb variance

The role of the bankroll is precisely to allow a strategy to survive poor runs.

If a bettor has a genuine Edge but risks too much capital on each bet, they can lose a large proportion of their bankroll before the statistical advantage has time to express itself.

That is exactly what Bankroll Management is designed to prevent.

With our reference:

Bankroll = €5,000

1U = €50

we have:

100 theoretical units.

That provides far more ability to absorb variance than a system where:

1U = €250

or:

5% of the bankroll.

Comparing 1% and 5%

Suppose we experience:

10 net losing units

With 1U = 1%

1U:

€50

10U loss:

-€500

Remaining bankroll:

€4,500

Loss:

-10%

With 1U = 5%

1U:

€250

10U loss:

-€2,500

Remaining bankroll:

€2,500

Loss:

-50%

Same strategy.

Same negative run.

Completely different consequences.

That is why Stake Sizing matters so much.

Variance becomes dangerous when Stakes are too large

Variance cannot be eliminated.

But its impact on the bankroll can be controlled.

If Stakes are too large, a statistically normal negative period can become catastrophic.

This can lead to:

• a major loss of capital;
• psychological pressure;
• irrational changes to Stakes;
• chasing;
• premature abandonment of a potentially profitable strategy.

Bankroll Management does not remove negative periods.

It increases the ability to survive them.

A 50% loss requires a 100% gain to recover

This is another reason why large Drawdowns should be avoided.

Suppose a bankroll starts at:

€5,000

After a:

50% loss

only:

€2,500

remains.

To return to €5,000, the bettor must then earn:

€2,500

which means:

+100% on the remaining capital.

Losses and recoveries are therefore not symmetrical.

Limiting Drawdowns is therefore essential.

Why increasing Stakes after a bad run is dangerous

After several losses, some bettors try to recover quickly.

They increase their Stakes.

That is usually the wrong response.

A previous losing sequence does not automatically make the next bet more likely to win.

If the Stake increases from:

1U

to:

3U

simply because the previous bets lost, the risk increases even though the Edge may not have changed at all.

This is chasing.

Variance then becomes much more dangerous.

A losing streak does not make the next bet more likely to win

This is often referred to as the Gambler's Fallacy.

After 5 or 6 losses, a bettor may think:

"The next one has to win."

That is false.

If each bet is a new opportunity with its own probability, previous losses do not guarantee the next outcome.

A bad run can continue.

The bankroll must be designed with that possibility in mind.

When should you start to worry?

There is no universal number.

A negative period must be compared with the expected statistical behaviour of the strategy.

You should examine factors such as:

• number of bets;
• loss in units;
• historical Drawdown;
• average odds;
• win rate;
• historical ROI;
• CLV;
• changes in liquidity;
• possible market changes.

Losing 5U may be completely normal.

Losing 30U may also be possible for certain strategies.

The important point is to compare the current Drawdown with the strategy's historical and statistical profile.

CLV can help distinguish variance from deterioration

Financial results are not the only information available.

Closing Line Value (CLV) can provide additional insight.

Suppose a strategy is currently down:

-12U

over a recent period.

But it continues to consistently obtain better odds than the market's Closing Odds.

That can be a meaningful signal.

The financial results have been negative, but the selection process still appears to be beating the market.

On the other hand, if:

• results turn negative;
• CLV deteriorates;
• ROI declines;
• the estimated Edge disappears;

then it becomes more reasonable to reassess the strategy.

Distinguishing between:

bad variance

and:

a genuine deterioration of the Edge

is essential.

A good process can produce a bad result

This is a fundamental idea in investing, trading and sports betting.

In the short term:

good process ≠ guaranteed good result

But over the long term, the objective is to build a process with positive Expected Value.

The bettor should therefore assess:

the quality of the process

before focusing only on:

recent results.

That includes analysing:

• the quality of the odds obtained;
• estimated Value;
• CLV;
• Stake discipline;
• betting volume;
• model stability.

How many bets are needed to know whether a strategy is profitable?

There is no universal number.

It depends on:

• the true Edge;
• variance;
• odds;
• market type;
• win rate;
• the dispersion of results.

A few dozen bets are usually far too few to support a strong conclusion.

Even several hundred bets can sometimes leave substantial statistical uncertainty.

The smaller the true Edge, the larger the sample generally needed to detect it clearly.

That is why performance should always be interpreted alongside:

sample size.

50 positive bets do not prove that a strategy is good

Suppose a strategy is:

+12U after 50 bets

That may be encouraging.

But it is not sufficient proof of a durable Edge.

A poor strategy can experience an excellent 50-bet run because of variance.

Conversely, a genuinely profitable strategy may show:

-6U after 50 bets

simply because of a bad run.

Time and volume gradually help separate:

luck

from:

skill.

Why 1,000 bets provide more information

As the sample grows, results generally become more informative.

A track record of:

1,000 bets

does not eliminate variance.

But it provides much more data for analysing:

• ROI;
• Drawdown;
• stability;
• odds;
• win rate;
• CLV;
• Flat Stakes performance.

A longer record also allows us to observe how the strategy behaved across different market periods.

Beware of recency bias

Recency bias means giving too much weight to the most recent events.

For example:

A Tipster has:

2,000 bets

with:

+80U

Then the most recent 50 bets show:

-10U

Some users may immediately decide that the Tipster has become poor.

But those 50 bets represent only:

2.5% of the total record.

That does not mean the recent period should be ignored.

But it must be placed in its statistical context.

The opposite mistake also exists

Recency bias works in both directions.

Suppose a weak strategy earns:

+15U over its latest 30 bets.

That does not automatically mean it has become excellent.

A very positive run can also be caused by variance.

Two symmetrical mistakes should be avoided:

abandoning a good strategy after a bad run

and:

trusting a bad strategy after a good run.

The psychological role of the bankroll

The bankroll is not only there to absorb losses mathematically.

It also plays a psychological role.

A run of:

-10U

is easier to manage if it represents:

-10% of the bankroll

than if it represents:

-50%.

The larger the loss becomes relative to available capital, the harder it is to remain rational.

A properly sized bankroll therefore protects:

capital

but also:

discipline.

Variance is part of the price of exploiting an Edge

An Edge does not produce a smooth return like a salary.

Returns are uncertain over time.

The bettor therefore accepts a form of volatility in exchange for positive long-term expectation.

That volatility is not an anomaly.

It is part of the process.

The objective is not:

to eliminate variance

because that is impossible.

The objective is:

to build a bankroll and Stake Sizing system capable of withstanding it.

What should you do during a bad run?

A negative period should not be ignored.

But it should be analysed methodically.

It can be useful to check:

• whether the strategy rules are still being followed;
• whether Stakes have changed;
• whether the markets have changed;
• whether the odds obtained remain competitive;
• whether CLV remains positive;
• whether the Drawdown is consistent with the historical profile;
• whether the betting volume is large enough to support a conclusion.

The objective is neither to panic nor to ignore warning signs.

The key is to distinguish:

normal variance

from:

structural deterioration.

Why Flat Stakes Summary (1U) is useful during a Drawdown

When a Tipster uses variable Stakes, it can be difficult to know whether a bad period comes from:

the selections

or:

Stake Sizing.

The Flat Stakes Summary (1U) recalculates every selection using exactly:

1U

This provides an additional perspective.

Suppose:

Recent actual result: -15U

but:

Flat Stakes Summary (1U): -5U

A significant part of the Drawdown probably came from Stake weighting.

On the other hand:

Actual result: -12U

Flat Stakes Summary (1U): -11U

suggests that the selections themselves experienced an unfavourable period.

Variance and diversification

Diversification can also influence the variance profile.

A strategy concentrated only on:

one league

one type of bet

or:

one market

may be more exposed to specific effects.

A coherent diversification across genuinely independent strategies can sometimes reduce overall volatility.

But simply adding more bets does not automatically create diversification.

If the bets are highly correlated, risk can remain high.

Be careful with correlated bets

Suppose you have five open bets on the same match or on strongly related outcomes.

Even if each bet is only:

1U

the real exposure may be greater than each individual Stake suggests.

For example:

5 bets × 1U = 5U of exposure

With:

1U = €50

that represents:

€250

or:

5% of the starting bankroll.

Bankroll Management should therefore also consider simultaneous exposure.

Is variance easier to accept when you understand the maths?

Usually, yes.

A bettor who expects to win every week will interpret a negative period as failure.

A bettor who understands probability knows that a winning strategy can experience:

Losing Streaks

Drawdowns

flat periods

without the Edge necessarily disappearing.

That understanding changes how results are evaluated.

The question becomes less:

"Did I make money this week?"

and more:

"Does my process still have a measurable advantage over a sufficiently large sample?"

Conclusion

A good bettor can lose for several weeks.

That is not a contradiction.

It is a normal consequence of variance.

A strategy with positive Expected Value can experience:

• several consecutive losses;
• negative weeks;
• several difficult months;
• significant Drawdowns;
• long periods of stagnation.

That does not automatically mean the Edge has disappeared.

In the short term, results can be dominated by randomness.

Over the long term, if the Edge is real and sufficiently stable, it should gradually become visible over a larger sample.

That is precisely why Bankroll Management is essential.

With our example:

Bankroll: €5,000

1U = €50

1U = 1% of the starting bankroll

a negative run is much easier to absorb than if Stakes represent 5% or 10% of capital.

The objective is not to prevent Drawdowns.

They are unavoidable.

The objective is:

not to risk so much on each bet that a statistically normal Drawdown destroys the bankroll.

You should also avoid judging a strategy only by its most recent results.

A serious analysis should include:

number of bets

ROI

Drawdown

odds

CLV

liquidity

track record duration

and:

Flat Stakes Summary (1U)

The key question is therefore not:

"Why did this strategy lose this week?"

but rather:

"Are the observed losses consistent with the normal variance of a strategy that still has an Edge?"

Understanding that distinction is one of the most important steps in moving from an emotional view of sports betting to a genuinely statistical and methodical approach.

FAQ: Bankroll and Variance in Sports Betting

Can a good bettor lose for several consecutive weeks?

Yes. Even a strategy with positive Expected Value can go through several negative weeks. Short-term results are heavily influenced by variance.

Does a losing streak mean the strategy no longer works?

Not necessarily. You need to analyse the number of bets, Drawdown, ROI, odds, CLV and the broader track record before reaching a conclusion.

How many consecutive losses can a good strategy experience?

There is no universal number. It depends mainly on win rate, average odds and total betting volume. High-odds strategies can experience especially long Losing Streaks.

What is variance in sports betting?

Variance describes the natural fluctuations of actual results around their mathematical expectation. It explains why a profitable long-term strategy can suffer significant short-term losses.

Why is a sufficiently large bankroll necessary?

A properly sized bankroll allows a bettor to absorb bad runs without putting the entire capital at risk. It gives the Edge enough statistical time to express itself.

Can a strategy with positive ROI still be negative after 100 bets?

Yes. Positive long-term ROI does not guarantee a positive result after 100 bets. Over a limited sample, variance can dominate the Edge.

How can you tell whether a bad run is variance or the disappearance of the Edge?

You need to analyse several indicators, including sample size, CLV, ROI, Drawdown, odds obtained, market evolution and whether the original process is still being followed. No single metric is usually enough.

Should you increase Stakes to recover after a losing streak?

No. Arbitrarily increasing Stakes to recover losses is chasing and significantly increases risk. A losing streak does not make the next bet more likely to win.

Why is CLV useful during a negative period?

Consistently positive CLV can indicate that the process is still obtaining favourable prices relative to the market even when recent financial results are negative. It therefore provides useful context beyond simple profit and loss.

Can Flat Stakes Summary (1U) help analyse a Drawdown?

Yes. It recalculates all selections with the same 1U Stake and helps distinguish the impact of selection quality from the impact of Stake Sizing.