Dice guide
Diamond Dice: what a different house edge does
"Diamond" names this page and does not mark a premium version of anything. The subject is a single dial in the dice formula, the house edge, and what happens to the numbers when it moves.
One target, five prices
Hold the target fixed: roll under 25.00, which wins on 2,500 of 10,000 results. A fair game pays 4.00x. Every percentage point of edge removes 1% of that fair payout. The table below shows the same bet priced five ways.
| House edge | Payout | EV per unit | Expected loss over 1,000 one-unit bets | SD of net over 1,000 bets |
|---|---|---|---|---|
| 0.5% | 3.9800x | -0.50% | 5.0 | 54.5 |
| 1% | 3.9600x | -1.00% | 10.0 | 54.2 |
| 2% | 3.9200x | -2.00% | 20.0 | 53.7 |
| 3% | 3.8800x | -3.00% | 30.0 | 53.1 |
| 5% | 3.8000x | -5.00% | 50.0 | 52.0 |
Model: 10,000 equally likely results; payout (1 - edge) x 100 / 25.
What the table says
The payout column moves only in the second decimal place, from 3.98x to 3.80x. On one bet the difference is barely visible. Over 1,000 bets the expected loss grows from 5 units to 50 units, a factor of ten. The edge is a rate, so its effect grows in step with the number of units wagered.
The spread barely moves. At every edge in the table, the standard deviation over 1,000 bets is close to 54 units. In a short session, luck swamps the edge. In a long one, the edge wins. One way to see the crossover: the expected loss equals one standard deviation after about (SD per bet / edge) squared bets. At a 1% edge that is 29,403 bets; at 5% it is 1,083.
The same comparison at other targets
| Win chance | Roll needed | Payout | EV per unit | SD per bet |
|---|---|---|---|---|
| 5% | under 5.00 | 19.4000x | -3.00% | 4.228 |
| 33% | under 33.00 | 2.9394x | -3.00% | 1.382 |
| 66% | under 66.00 | 1.4697x | -3.00% | 0.696 |
| 95% | under 95.00 | 1.0211x | -3.00% | 0.223 |
At a 3% edge the EV column reads -3.00% on every row, just as it reads -1% at a 1% edge. That confirms the edge belongs to the pricing rule, not to any one target. Choosing a target cannot lower it. Only the rule itself, the number in (1 - e), sets the long-run cost.