Explanations
tilttek/dice-jailStatistics\Explanation::explain() writes a FitResult for the table, in a Language — English or French;
Language::ofLocale() reads fr, fr_CA or fr-CA — and for an Origin: Physical, a die thrown at the table,
or Digital, rolls drawn by software.
new Explanation()->explain($result, Language::French, Origin::Digital);
| Verdict | English |
|---|---|
| Not enough rolls | Not enough rolls to judge this die yet: 29 recorded, 30 needed. Until then, a streak is only chance. |
| Nothing unusual | Nothing unusual in 30 rolls: a fair die spreads its faces at least this unevenly 8.2% of the time (p = 0.0821). |
| Somewhat unusual | Somewhat unusual in 30 rolls: a fair die spreads its faces this unevenly about once in 181 series (p = 0.0055). It proves nothing; roll it again and judge the new rolls alone. |
| Very unusual, physical | Very unusual in 30 rolls: … fewer than once in 1,000 series (p = 0.0004). It is a reason to look at the die — its edges, its balance, how it is thrown — not proof that it is loaded. |
| Very unusual, digital | … Test new rolls before suspecting anything: chance remains the likeliest cause. |
The average roll follows, against the fair one: « Average roll: 1.67, against 2.00 for a fair die. » For digital dice, a last sentence says that the generator serves every die alike and that the jail is only a game.
The French sentences say the same, with French typography: a comma for decimals, a narrow no-break space between
thousands, a no-break space before « : » and « % », a narrow one before « ; ». Explanation::pValue() writes a
p-value alone: p = 0.0407, or p < 0.0001 below it.
What the sentences never say: that a die is loaded, cursed or cheating; that a member cheats; that a p-value is the chance of anything about the die.