When to Cash Out: The Decision That Changes Everything Except Your Return

Crash games, ladder games and anything with a cash-out button all reduce to the same question asked repeatedly: take what you have, or go one more? Every guide to when to cash out on a crash game will give you a number — stop at 2×, stop at 1.5×, use auto-cash-out at 1.3×. Almost none of them tell you the thing that makes those numbers make sense, which is that on a properly built ladder, every one of them is worth exactly the same.

Why every stopping point pays the same

Take a ladder where each step survives with a flat 90% chance. To keep the return constant, the multiplier at each rung has to be the target return divided by the chance of reaching that rung. That is precisely how Balloon Pump is built: one pump pays 1.0555× and you reach it 90.0% of the time; ten pumps pay 2.7245× and you reach them 34.9% of the time; twenty pumps pay 7.8140× and you reach them 12.2% of the time.

Multiply each pair and you get the same number every time — 95% of your stake. Stopping at the first rung, the tenth or the twentieth has identical expected value. The cash-out button is not an expected-value decision at all.

So what are you actually choosing?

Variance — the shape of your results, not their average. And that is a real choice with real consequences, just not the one people think they are making.

• Stopping early gives you many small wins and rare losses. A session looks calm. Your balance drifts down slowly, at the rate of the house edge, and rarely surprises you.
• Stopping late gives you many losses and occasional large wins. The same drift, arrived at through a much bumpier road, and with a far higher chance of a short session ending well above or far below where it started.

If someone tells you their cash-out point is optimal, the honest question is "optimal for what?" For expected return, nothing is optimal — they are all equal. For staying in a session longer on a fixed budget, stop early. For having any chance at a large result before your budget runs out, stop late. Those are different goals and they have different answers.

The memoryless trap

Where a game's risk per step is flat, the round has no memory. A balloon that has survived fourteen pumps is exactly as likely to survive the fifteenth as a fresh one is to survive its first: 90%, always.

Two very common instincts contradict this, and both cost money:

• "It is due to pop." It is not. Nothing accumulates. The next step carries the risk the next step has always carried.
• "It is on a run, ride it." Also no. A long run is a run that happened, not a property the round has acquired.

The reason this is worth stating plainly is that a game which does accelerate its risk — where each step is more dangerous than the last — feels identical from the outside. The only way to tell is to read the published rungs. A ladder game that will not show you its survival chance per step is asking you to guess at the one number the decision depends on.

Both instincts have names — the gambler's fallacy and the hot hand fallacy — and we test them against a quarter of a million simulated rounds in does the balloon remember?

When the cash-out decision does have a right answer

Three cases, and they are worth knowing because they are the exceptions that prove the rule.

1. When you are racing someone

In a duel the pot is winner-take-all, so a cautious loss pays exactly what a reckless one does — nothing. That destroys the symmetry immediately. In PvP Minefield, if you are behind and your opponent has stopped, you must keep going; if you are ahead and they have stopped, you must stop. Here the cash-out choice has a genuinely correct answer, and getting it wrong costs real money.

2. When there is a cap

A maximum payout truncates the top of the ladder. Once the rung you are climbing towards would pay more than the cap allows, the reward stops growing while the risk does not, and the expected value of that step falls below every step beneath it. Any rung near a win cap is strictly worse than stopping.

3. When the game charges per round rather than per step

Where a flat fee or a rake is taken once per round, a longer round spreads that cost over more play. It does not change the ladder, but it does mean a format that lets you get more game out of one pot is cheaper per decision.

What about auto-cash-out?

Setting a target in advance changes nothing about your expected return. It does change two things worth having: it removes the reaction-time question entirely, and it removes you from the decision at the exact moment you are least equipped to make it — mid-round, watching a number climb. If you have decided on a variance profile, automating it is how you actually stick to it.

Where a game resolves one tap at a time rather than on a running clock, as ours does, the reaction-time problem does not arise in the first place. The discipline problem still does.

The checklist

1. Find the survival chance per step. Flat, or accelerating? If the game will not say, that is your answer about the game.
2. Check the rungs against it. Multiplier times chance of reaching it should be the same constant at every rung. If it drifts downward as you climb, the top of the ladder is worse than the bottom and the game is not telling you.
3. Find the win cap and treat every rung near it as off the table.
4. Pick your variance and automate it, rather than deciding again every round.
5. In a duel, throw all of the above out and play the opponent.

The full ladder, with the survival chance beside every rung, is on the Balloon Pump guide. The RTP index shows how the same 95% is reached by every other game we run.