Most players with a genuine edge over a game carry a quiet frustration: the bankroll does not climb the way the math on paper says it should. Sessions end short. Recovery from downswings takes longer than expected. The numbers look right but the results feel wrong. The usual explanation is bad luck, and sometimes that is partly true. But there is a structural force grinding away at growth that has nothing to do with luck and everything to do with a relationship most players never examine directly: the interaction between edge and variance in determining actual, realized growth rate.
Expected Value Is Not Growth Rate
Expected value tells you the average outcome of a bet, weighted by probability. If you have a 1% edge on a $100 wager, your EV is $1. Play 1,000 hands and you expect $1,000. Clean, linear, satisfying. The problem is that expected value is a long-run average, not a growth trajectory. What actually accumulates in your bankroll is not EV—it is geometric return, and those two things are not the same.
Geometric return accounts for the fact that you are not betting a fixed dollar amount abstracted from a limitless supply of money. You are betting a fraction of a finite bankroll. When that bankroll shrinks due to variance, your subsequent bets are smaller in absolute terms, which means winning back what you lost requires a higher percentage gain than the loss that preceded it. Lose 20% of your bankroll and you need a 25% gain to get back to even. This asymmetry is not dramatic bet-to-bet, but it accumulates relentlessly over a long playing career.
The Drag Formula
There is a precise expression for this. The long-run geometric growth rate of a bankroll is approximately equal to expected value per bet minus one-half of variance per bet—expressed as fractions of the bankroll. Written concisely: g ≈ μ − (σ²/2), where μ is the mean return per unit and σ² is the variance.
That subtracted term is the drag. It is not a metaphor or an approximation of bad luck. It is a structural reduction in your realized growth rate that exists even in a perfectly fair sample. And crucially, it scales with variance—not with edge. A player in a high-variance game suffers more drag than a player with an identical edge in a low-variance game. The edge is the same. The growth rate is not.
Consider two players, each with a 1.5% edge on $100 bets. Player A is in a low-variance game where the standard deviation per bet is $90. Player B is in a high-variance game with a standard deviation of $250 per bet. The drag term for Player A is (8,100/2) divided by relevant bankroll fractions—small. For Player B it is (62,500/2) scaled similarly—roughly seven times larger. Both players have the same edge. Player B's bankroll grows materially slower, and in a smaller bankroll scenario, may not grow consistently at all.
What This Means in Practice
The drag effect has three concrete implications that change how a skilled player should approach their game.
First, variance selection is a strategic decision, not a preference. Many players choose games or bet structures based on excitement level or payout format without recognizing that higher variance directly penalizes bankroll growth independent of edge. If you have a choice between two games offering the same edge, the lower-variance game compounds your advantage more efficiently. This is not a trivial difference over thousands of sessions.
Second, bet sizing affects drag directly. Because drag is proportional to variance as a fraction of bankroll, betting a larger fraction of your bankroll amplifies drag even if it also amplifies expected return. This is the mathematical underpinning of why overbetting is destructive even when your edge is real. You are not just increasing risk—you are structurally reducing your geometric growth rate. The optimal bet size from a growth perspective balances edge against the drag penalty, which is exactly what Kelly-style sizing calculates.
Third, interpreting session results requires accounting for drag. If you expect to earn $500 in a session based purely on edge calculations and you earn $320, the instinct is to attribute $180 to variance. Sometimes that is correct. But part of that gap may be structural—the drag operating exactly as the math predicts. Players who do not understand this attribute normal drag to bad runs and respond by adjusting strategy unnecessarily, which introduces a second layer of error on top of the first.
The Compounding Timeline
Drag is particularly important over long time horizons because geometric effects compound. A player with a 2% edge and moderate variance might project their bankroll doubling in a certain number of sessions using linear EV math. Accounting for drag, that timeline extends—sometimes significantly. In high-variance games with aggressive bet sizing, the drag can be large enough that a player with a real edge barely grows their bankroll at all, and experiences extended periods of decline that are statistically expected rather than anomalous.
This is not a discouragement from playing positive-edge games. It is an argument for calculating the correct growth expectation from the start. A player who knows their realistic geometric growth rate does not panic during the flat stretches or over-correct after downswings. They recognize the pattern as the expected behavior of a system operating correctly.
The Practical Calculation
You do not need advanced software to estimate your drag. You need three numbers: your edge per bet expressed as a decimal, your standard deviation per bet, and your average bet size relative to your bankroll. Square the standard deviation, divide by two, and express the result as a fraction of your bankroll. That is your drag term. Subtract it from your edge fraction. The result is your approximate geometric growth rate per bet.
If that number surprises you—if it is lower than you assumed or even negative despite a positive edge—you now have specific, actionable information: reduce bet size, seek lower-variance game formats, or revisit whether your edge estimate is accurate. Any of those adjustments is more useful than the vague sense that variance is getting in the way.
Edge is necessary for long-run profitability. It is not sufficient for understanding how fast, or whether, your bankroll actually grows. Drag is the missing term, and filling it in makes the math honest.