From Simulated Wins to a Break-Even Payout
A game can be entertaining and still be priced so that players lose money on average. To assess its price, we need to know how often the game pays out and how much it pays. When the win probability is difficult or tedious to calculate exactly, a simulation can estimate it: repeatedly reproduce the game, count the wins, and use their proportion as an estimate of the probability of winning.
That estimated probability can then be used to find a break-even payout: the payout that makes the player's expected net winnings equal to zero, according to the probability estimate. This is a statement about an average over many plays under the model. It does not guarantee that a player will break even, or even win, in a particular play.
Here, “gross payout” means all the money the player receives from the game on a win; the entry cost is subtracted separately. If a game instead advertises a payout that includes returning the entry cost, make sure to use a consistent definition before calculating. A simulation estimates \(p\); it does not change the meaning of the payout or cost.
Why the Break-Even Formula Works
Let \(I\) equal 1 if a play wins and 0 otherwise. The net winnings \(Y\) are \(RI-c\): the player receives \(R\) when \(I=1\), and pays \(c\) in either outcome. Since \(P(I=1)=p\), the expected value of \(I\) is \(p\). By linearity of expectation, the expected net winnings are \(R\) times the win probability minus the cost.
Break-even means \(E(Y)=0\), so \(Rp-c=0\). If \(p>0\), dividing by \(p\) gives \(R=c/p\). If \(p=0\) and the cost is positive, no finite win-only payout can break even: a player never wins, so expected net winnings are always \(-c\).
This derivation uses a probability for the game's win event and a payout and cost that do not vary across plays. It describes expected value under the specified game model. It does not imply that actual winnings will equal expected winnings over a small number of plays.
A Simulation Estimates the Win Probability
A simulation trial must represent one complete play of the game. For each trial, apply the game rules as specified, record whether the player wins, and then reset or restore the game to its starting conditions as needed before the next trial. In games involving draws without replacement, the draws within a trial remain without replacement; restoring the cards between trials allows each simulated play to begin afresh.
If \(N\) simulated plays include \(W\) wins, the estimated win probability is the simulated proportion of wins, \(\hat{p}=W/N\). The simulation uses \(\hat{p}\) in place of the unknown or inconvenient-to-calculate \(p\), giving the estimated break-even payout \(c/\hat{p}\), provided \(\hat{p}>0\).
Write down the random events, the win condition, the entry cost, and what is paid for each possible outcome.
Generate outcomes using a method that follows the game rules. Record a win or loss for every complete play.
Compute \(\hat{p}=W/N\), where \(W\) is the number of simulated wins and \(N\) is the number of simulated plays.
For a win-only payout, compute \(c/\hat{p}\). State that this is an estimated break-even payout in expected value, not a guaranteed outcome.
Worked Examples
Worked Example: Estimating a Win-Only Break-Even Payout
A hypothetical game costs $3.00 to play. A player shuffles 10 distinct cards, 3 of which are marked, and draws 3 cards without replacement. The player wins if exactly 2 of the 3 drawn cards are marked. To estimate the break-even payout, a simulation repeats the complete game 10,000 times, restoring and reshuffling the cards between plays. Suppose 1,764 simulated plays are wins.
The trial matches the game: cards are drawn without replacement within each play, and the cards are restored before the next play. The simulation records one outcome per complete play, so the estimated win probability is
Using the entry cost \(c=\$3.00\), the estimated break-even gross payout is
At this estimated payout, estimated expected net winnings are approximately \(0.1764(\$17.01)-\$3.00=\$0.000564\) per play, effectively zero at the precision of cents. This does not mean that every player receives $17.01, or that a player will break even after a play. A player either wins the payout or loses the entry cost; the calculation describes the average net winnings implied by the estimated probability over many plays.
The estimated probability is based on a finite simulation. For example, the approximate standard error of the simulated proportion is \(\sqrt{\hat{p}(1-\hat{p})/N}=\sqrt{0.1764(0.8236)/10000}\approx0.00381\). This indicates that some variation in the simulated proportion is expected across runs of this size. Because the break-even payout divides the cost by the estimated probability, a different simulated proportion would produce a different estimated payout.
Worked Example: Including a Consolation Payout
In a different hypothetical game, a player pays $3.00. A winning player receives a gross payout \(R\), while a losing player receives a $0.75 consolation payout. In 10,000 simulated plays, 2,200 are wins, so the estimated probability of winning is
The payout on a win must now be balanced against both the cost and the consolation money paid to losing players. A win contributes \(R\) to the player's gross receipt, while a loss contributes $0.75. Subtract the $3.00 cost in either outcome. The estimated expected net winnings are therefore
Set this expression to zero to find the estimated break-even winning payout:
At a $10.98 winning payout, the estimated expected net winnings are \(0.22(\$10.98)+0.78(\$0.75)-\$3.00=\$0.0006\) per play, approximately zero. Using the win-only formula \(c/\hat{p}\) here would ignore the $0.75 received on losses and would give an incorrect break-even payout. All outcome payouts must be included in the expected value.
Worked Example: Evaluating an Offered Payout
A hypothetical game costs $4.00 and pays $25.00 on a win and nothing on a loss. A simulation of 2,000 complete plays produces 312 wins. Estimate the win probability and determine whether the offered payout is above or below the estimated break-even amount.
The simulated proportion of wins is
The estimated break-even payout is
The offered payout, $25.00, is below this estimated break-even amount. Equivalently, use the simulated probability to estimate expected net winnings directly:
Under the simulated estimate, the player's average net winnings are about -$0.10 per play. This is not a prediction that every player loses exactly 10 cents, or that every individual play is a loss. It is an average-value comparison based on the simulated win proportion. If the payout were $26.00 instead, estimated expected net winnings would be \(0.156(\$26.00)-\$4.00=\$0.056\), or about 5.6 cents per play.
What “Fair” Does and Does Not Mean
In this setting, “fair” has a specific, limited meaning: the payout makes expected net winnings zero under a stated probability model. A game can be fair in this expected-value sense while most individual plays lose. For instance, a relatively rare large payout can balance a frequent entry-cost loss when averaged over many plays. Fairness here does not mean that wins and losses are equally likely, that every player receives the same result, or that the game has no risk.
A payout calculated from a simulation is an estimate because \(\hat{p}\) varies from run to run. Increasing the number of simulated plays usually reduces the typical random variation in the estimated proportion, but no finite run guarantees the exact probability. The payout estimate also depends on whether the simulation correctly represents the game: an incorrect win condition, a failure to restore cards between plays, or a random generator that does not represent the intended outcomes can make the estimate misleading.
If a simulation produces no wins, then \(\hat{p}=0\), and the calculation \(c/\hat{p}\) is undefined. This does not establish that a win is impossible. It means only that the simulation observed no wins; the run may be too small to provide a useful estimate for a rare event. The appropriate response is to inspect the model and consider whether more simulation is needed, rather than to report a finite break-even payout from division by zero.
Common Mistake: Treating an Estimate as a Guarantee
A common error is to say, “The simulation found a fair payout of $17.01, so a player who pays $3.00 is guaranteed to get $17.01 back often enough to break even.” That reasoning confuses an estimated average with an individual outcome. In the card-game example, the player receives $17.01 only when the game is won and loses $3.00 otherwise. The $17.01 figure is a break-even estimate based on the simulated proportion, not a promise of a refund, a guaranteed long-run average for a particular player, or proof that the true win probability is exactly 0.1764.
A careful conclusion gives both the calculation and its qualification: “The simulation estimated a win probability of 0.1764, implying an estimated break-even gross payout of about $17.01 for a $3.00 entry cost. The estimate is subject to simulation error, and an individual play can still win or lose.” This wording distinguishes what the model estimates from what the game guarantees.
Check Your Understanding
Use the definitions and calculations in this tutorial to answer the following questions.
- A game costs $5.00, pays $20.00 on a win, and pays nothing on a loss. If the estimated win probability is 0.24, what is the estimated expected net winnings per play?
- A simulation records 480 wins in 2,000 plays. What is the estimated win probability? For an entry cost of $6.00 and no payout on a loss, what is the estimated break-even gross payout?
- A losing player receives $1.00 in a game that costs $4.00. In addition to the win probability, what information is needed to calculate the break-even winning payout?
- Why does a simulated break-even payout not guarantee that a player breaks even on one play?
- A simulation records no wins. Why is it not valid to conclude from this alone that no finite break-even payout exists for the actual game?