Spunk Odds Reference

Dice guide

Ice Dice: a seeded simulation, read line by line

The name Ice belongs to this page only; it describes no separate game. Here we run a computer simulation of number dice and read the output carefully, to see what independent rolls can look like and what they cannot tell you.

What was simulated

The simulation uses Python's built-in pseudo-random generator with a fixed seed, 7711, so anyone can repeat it and get the same numbers. Each simulated roll draws one of 10,000 equally likely values. The bet is roll under 66.00, a 66% win chance, which pays 1.5000x at a 1% edge. Six sessions of 150 one-unit rolls were run.

Simulation: roll under 66.00, 1% edge, six sessions of 150 rolls (seed 7711)
SessionRounds wonNet (units)Lowest pointLongest losing run
197-4.50-14.504
21046.00-3.504
31069.000.003
41046.00-4.003
595-7.50-7.504
695-7.50-7.503

Seeded pseudo-random simulation (Python random.Random, seed 7711), 6 sessions of 150 one-unit rounds. Average session result 0.25 units. This is a simulation, not a record of real play; a different seed gives different numbers.

How to read each column

Rounds won
The count of winning rolls out of 150. The expected count is 99, with a standard deviation of 5.8.
Net (units)
Wins times 0.5000, minus losses. The expected net is -1.50 units per session.
Lowest point
The deepest the running total fell during the session. It shows how much a player would have needed to keep going.
Longest losing run
The longest stretch of consecutive losses. At a 34% loss chance, runs of four or five are routine over 150 rolls.

What the simulation shows, and what it does not

The six sessions spread around the expected value rather than sitting on it. Some end ahead, some behind. The standard deviation of one session's net is 8.70 units, about 6 times larger than the expected loss, so a session in profit is not unusual.

A simulation is not a record of real play. It also proves nothing about any service. It only illustrates the model. Change the seed and every row changes, while the average behavior stays the same.

A longer run

With the same seed, a run of 100,000 rolls gives an average result of -0.0101 units per roll. The model's exact value is -0.0100. The standard error of a 100,000-roll average is 0.0022, so a result within about two of those of -0.01 is what the model predicts.

Why use a fixed seed

A fixed seed makes the output reproducible. That matters for teaching: the numbers on this page will match what you get if you run the same code. Real services do not use a fixed public seed, of course. They use the commit-and-reveal method described on the hash fairness guide, so that results are unpredictable in advance but checkable afterwards.