Spunk Odds Reference

Dice guide

Gold Dice: how long until the first win

Gold is a label this page carries; the content is about number dice in general. The question is about time rather than money: at a 10% chance, how many rolls pass before the first win?

The geometric distribution

Pick a roll-under target of 10.00, which wins on 1,000 of 10,000 results. Each roll is a fresh 10% trial. The roll on which the first win arrives follows a geometric distribution. The chance that the first win is roll k is 0.9 to the power k - 1, times 0.1. The chance of at least one win by roll k is 1 - 0.9 to the power k.

Roll under 10.00 (10% chance): when the first win arrives
Round kP(first win exactly at k)P(first win by round k)
110.000%10.00%
38.100%27.10%
75.314%52.17%
103.874%65.13%
221.094%90.15%
440.108%99.03%

Mean, median and the tail

The mean wait is 1 / 0.1 = 10 rolls. The median is shorter: by roll 7 more than half of all sessions have seen a win. The gap between the two comes from the long right tail. One session in ten is still waiting after 21 rolls, and one in a hundred is still waiting after 43.

The most likely single roll for the first win is roll 1, with 10.0%. Each later roll is a bit less likely than the one before. That surprises people, who expect the first win to cluster around roll 10. It does not cluster; it spreads out, with a slow decline.

No memory

The geometric distribution has a property called memorylessness. If you have already waited 15 rolls without a win, the expected number of further rolls is still 10. The 15 rolls already spent do not shorten the wait ahead. Each roll is a new 10% trial, whatever came before.

What the wait costs

At a 1% edge, the 10% bet pays 9.90x. If the first win comes on roll k, the player has staked k units and received 9.9. The session is ahead at that moment only if k is 9 or less. The chance of that is 61.26%. The rest of the time, the first win arrives after the stakes already exceed what it pays back.

Other chances, same shape

In each case the 99% point lands between about 4 and 4.6 times the mean, because the tail keeps a similar relative shape. Lower chances stretch the whole curve, which is worth knowing before picking a small target.