1. The Probability Density Function of Crash Multipliers
To reason rigorously about Expected Value, we must first establish the probability distribution governing crash game outcomes. In a Provably Fair crash game with configured house edge E (where 0 < E < 1, typically 0.03), the multiplier X is a continuous random variable on the support [1.00, +∞).
The survival function (complementary CDF) takes the following closed form:
P(X >= k) = (1 - E) / k for k >= 1.00
This describes a truncated reciprocal distribution with a point mass of probability E at exactly X = 1.00 (instant crash). For the continuous portion above 1.00x, the probability density function (PDF) is:
f(x) = (1 - E) / x² for x > 1.00
We verify this is a proper distribution: the integral of f(x) from 1 to ∞ equals (1 - E), plus the point mass E at 1.00, summing to 1.
2. Formal Derivation of Expected Value
Consider a player who places a bet of size B = 1 with auto-cashout at multiplier k (where k > 1.00). The payoff structure is binary:
- Win: If
X >= k, net profit =(k - 1) * B - Loss: If
X < k, the player losesB
The Expected Value:
EV(k) = P(X >= k) × (k - 1) - P(X < k) × 1
= [(1 - E) / k] × (k - 1) - [1 - (1 - E) / k] × 1
= [(1-E)(k-1)/k] - [(k - 1 + E)/k]
= [(1-E)(k-1) - k + 1 - E] / k
= [k - 1 - Ek + E - k + 1 - E] / k
= -Ek / k
= -E
The variable k cancels completely from the equation.
This is a remarkable result: the Expected Value per unit wagered is -E for every possible cashout multiplier. Whether you set auto-cashout at 1.01x, 2.00x, 10.00x, or 1000.00x, the mathematical expectation is identically -E.
For a standard 3% house edge: EV = -0.03 per unit wagered, unconditionally.
3. Variance Decomposition: Why Strategies Feel Different
If EV is constant across all targets, why do low-multiplier and high-multiplier strategies produce dramatically different session outcomes? The answer lies in variance.
Evaluating for representative targets with E = 0.03:
| Target k | Win Prob | Variance | Std Dev | EV |
|---|---|---|---|---|
| 1.10x | 88.2% | 0.104 | 0.322 | -0.03 |
| 2.00x | 48.5% | 0.999 | 0.999 | -0.03 |
| 10.00x | 9.7% | 8.729 | 2.955 | -0.03 |
| 50.00x | 1.94% | 47.53 | 6.895 | -0.03 |
The rightmost column is invariant. Variance grows monotonically with the target, creating a powerful cognitive illusion: a 1.10x player experiences long winning runs punctuated by occasional wipeouts, while a 50.00x player endures long losing droughts with rare large payouts. Both converge to -3%, but their subjective experience differs radically.
4. The Kelly Criterion and Optimal Sizing
The Kelly Criterion (John L. Kelly Jr., Bell Labs, 1956) prescribes the fraction of bankroll to wager in order to maximize long-term geometric growth rate. For a crash game at target k with house edge E:
f* = ((k-1) × (1-E)/k - (1-(1-E)/k)) / (k-1) = -E / (k - 1)
The Kelly fraction is negative for all k > 1 when E > 0. A negative Kelly fraction means the mathematically optimal strategy is to wager nothing. There is no target multiplier, no betting pattern, and no progression system that transforms a negative-EV game into a positive-EV proposition.
5. Implications for Rational Play
The framework above permits one conclusion: crash games are entertainment products with a guaranteed long-term cost, not investment vehicles. The house edge acts as a deterministic tax on every wager.
Explore the numerical implications with our interactive Expected Value & RTP Calculator.