The Math of Variance: Why Players With an Edge Still Go Broke

An edge is a statement about the long run. Variance decides whether you are still playing when the long run arrives. This is the part of gambling maths that people who are good at the game most often underestimate.

Two numbers, not one

Every repeated bet is described by two numbers:

• Expected value (μ): the average result per bet.
• Standard deviation (σ): how far a single result typically lands from that average.

Over n bets the expectation grows in proportion to n, but the standard deviation only grows with √n. That is why edges win eventually, and it is also why "eventually" can take a very long time.

A worked example: the card counter

In our ten-million-hand simulation, a Hi-Lo counter with a 1-to-12 unit spread had:

• μ = +0.005 units per hand (half a unit per hundred hands)
• σ = 2.51 units per hand

A player with a genuine edge still has close to even odds of being behind after a thousand hands, which is about ten hours of play. It takes about a million hands before the result is reliably positive. At 100 hands an hour, that is 10,000 hours.

(The last column uses the normal approximation: the chance of being behind is Φ(−μ√n / σ).)

Risk of ruin

Risk of ruin is the probability that a bankroll hits zero before the edge carries it clear. For a small edge and many bets there is a good approximation:

Using the counter's numbers above:

With a €10 unit, a counter needs around €20,000 set aside to have a 96% chance of never going broke. The edge is the same at every size. What changes is only whether the player survives long enough to collect it.

Kelly: how much to bet

The Kelly criterion answers the related question of how large each bet should be to grow a bankroll as fast as possible without risking ruin. For a simple even-money bet that you win with probability p:

A 51% coin flip at even money has a Kelly fraction of 2%: bet 2% of your current bankroll each time. Betting more than twice the Kelly fraction produces negative long-run growth even though every single bet has positive expectation. That result surprises most people.

For a negative-expectation bet, which is every standard casino bet, the Kelly fraction is zero. The formula agrees with the obvious advice.

The lesson

Expected value says where you are heading. Variance says how bumpy the road is, and bankroll size decides whether you get there. This applies beyond gambling: to trading, to startups, and to any repeated decision under uncertainty.