Understanding Variance: When Good Bets Lose

Here's a scenario that breaks most bettors: you have a genuine 5% edge. You place 200 bets. You lose money.

This is not only possible — it's common. It's not bad luck in any meaningful sense. It's variance, and it's a fundamental feature of probabilistic outcomes. Understanding it changes how you think about results, bankroll management, and what it actually means to be a "good bettor."

What Variance Actually Means

Variance is the spread of possible outcomes around an expected value. Even if your expected value is positive, individual outcomes are still random. Enough randomness, and you can lose for a long time before your edge expresses itself.

Consider 100 bets with 5% ROI at even odds (2.00):

True win rate: 52.5%
Expected profit per $100 wagered: $5
Standard deviation over 100 bets: ≈ $100

The standard deviation is twenty times the expected profit. A single standard deviation swing (which happens about 68% of the time) covers a range from +$105 to −$95 profit.

A coin flipper and a skilled bettor with 5% ROI look essentially identical over 100 bets. You cannot tell them apart.

Downswings Are Normal, Not Evidence of Failure

With a genuine 5% edge (52.5% win rate at even odds), here are rough probabilities of experiencing downswings:

20-bet losing streak: happens roughly once every 400 sequences
15% drawdown: occurs about 35% of the time over a 500-bet sample
Being negative after 200 bets: happens about 15% of the time

These are not disasters. They're expected features of the probabilistic landscape. The bettor who quits after a 15% drawdown, convinced their edge has disappeared, is often quitting right before their results converge on the true expected value.

The Minimum Sample Problem

How many bets do you need before you can distinguish skill from luck?

For a 5% ROI (a solid genuine edge), you need roughly 2,000–3,000 bets to achieve 95% statistical confidence that your results are not luck. For smaller edges (1–2%), the number is much larger.

Most bettors evaluate their edge over 50–200 bets. This is essentially noise. You can't conclude much of anything from a 200-bet sample, in either direction.

Bankroll Management and Variance

This is why bankroll management matters. Your bet sizes need to be small enough relative to your bankroll that you can survive the inevitable variance before your edge materializes.

The Kelly Criterion naturally accounts for this: it prescribes smaller bets when your edge is uncertain or small, which builds in variance protection.

A rough rule of thumb for recreational bettors: no single bet should exceed 2-3% of your total bankroll, and even Kelly-sized bets are often fractioned to 25-50% to reduce variance further.

How to Think About Results

When your results are worse than expected:

Check if the underlying logic of your edge has changed
Check if the market has adapted to your approach (lines moving against you faster)
Consider whether you're getting realistic odds (line shopping quality)
Check your bet sizing hasn't drifted up on losing streaks (classic mistake)

If none of these have changed, you're probably in the variance. The discomfort is real; the conclusion that your edge is gone is usually wrong.

When your results are better than expected:

Be equally skeptical
Running hot doesn't validate your model any more than running cold invalidates it

The results that matter are the ones over thousands of bets, measured against a consistent, pre-defined process.

The Uncomfortable Summary

Variance means that good bettors lose money for extended periods, and bad bettors make money for extended periods. Short-term results are almost meaningless as signals of edge.

This has an important corollary: most people claiming to have a betting system are in a period of positive variance, not genuine edge. It will eventually revert. The system will "stop working," at which point they'll find a new one.

The real signal is the process, the models, and long-run closing line value — not whether you were up last month.