Spunk Odds Reference

Coin flip guide

Crystal Coin Flip: how likely is a bankroll to run dry?

The Crystal name on this page is cosmetic. It refers to no special coin. The topic is bankroll risk: how often a fixed amount of money runs out when it is staked one unit at a time.

The question

Suppose you bring a set number of units and stake one unit per flip until you have staked a fixed number of flips or have nothing left. What is the chance you end at zero? The answer depends on the starting amount, the number of flips, and the edge.

How the model handles the payout

The exact calculation on this site tracks whole-unit balances, so it needs a win to return a whole number of units. A 1.98x payout does not fit. Two versions get around this honestly:

Bust probability: win chance 50.00%, win returns 2 units total (exact calculation)
Starting bankroll (units)P(bust within 300 one-unit bets)
577.29%
1538.65%
402.10%
Bust probability: win chance 49.50%, win returns 2 units total (exact calculation)
Starting bankroll (units)P(bust within 300 one-unit bets)
581.01%
1544.58%
403.09%

Reading the tables

Even with no edge at all, a 5-unit bankroll is lost within 300 flips 77.29% of the time. Small reserves fail through ordinary variance; no unfairness is needed. With the edge moved into the win chance, that rises to 81.01%. The edge matters most for the large bankroll: at 40 units the risk goes from 2.10% to 3.09%.

Longer play

Ruin keeps growing with time. For 15 units the edge-free risk is 38.65% after 300 flips and 69.85% after 1,500. With the 49.5% model it reaches 79.60% after 1,500 flips. In the edge-free game, play without end loses any finite bankroll eventually; with an edge it simply happens sooner.

Practical takeaways

  1. Decide the amount you can afford to lose before starting, and treat it as spent.
  2. Smaller stakes relative to the bankroll lower the chance of hitting zero within a session, but not the expected cost per unit staked.
  3. A session limit in flips is as useful as a money limit, because risk grows with time.