ᱵᱷᱨᱚᱢ ᱠᱚ · ᱵᱷᱨᱚᱢ ᱠᱚ · 6 ᱢᱤᱱᱤᱴ

ᱵᱷᱨᱚᱢ: ᱮᱞ ᱦᱤᱡᱩᱜ ᱞᱟᱹᱠᱛᱤ

“ᱜᱮᱞ ᱮᱭᱟᱭ 150 ᱜᱷᱩᱨᱟᱹᱣ ᱨᱮ ᱵᱟᱝ ᱦᱮᱡ ᱟᱠᱟᱱᱟ᱾ ᱱᱤᱛ ᱦᱮᱡ ᱞᱟᱹᱠᱛᱤ᱾” ᱵᱟᱝ᱾

ᱠᱷᱟᱴᱚ ᱛᱮᱞᱟ

On independent uniform single-zeroᱪᱟᱠᱟᱡᱤᱨᱚᱢᱤᱫᱜᱮ ᱦᱟᱹᱨᱭᱟᱹᱲ ᱠᱷᱟᱱᱟ ᱡᱟᱦᱟᱸ ᱥᱚᱢᱟᱱ ᱴᱟᱠᱟ, ᱰᱚᱡᱮᱱ ᱟᱨᱵᱟᱝ ᱠᱚᱞᱚᱢ ᱵᱟᱡᱤ ᱵᱟᱭ ᱠᱚᱵᱷᱟᱨᱟ; ᱠᱮᱥᱤᱱᱳ ᱵᱟᱲᱛᱤ ᱨᱮᱭᱟᱜ ᱢᱩᱞ᱾ ᱮᱴᱟᱜ ᱧᱩᱛᱩᱢ: ᱥᱤᱝᱜᱚᱞ ᱡᱤᱨᱚ, 0᱾ᱯᱩᱨᱟᱹ ᱯᱟᱲᱦᱟᱣ → spins, a number’s chance on the next spin remains 1/37, however long it has been absent. Calling it “due” mistakes a long-run proportion for a promise about the next result.

ᱢᱚᱰᱮᱞ
ᱟᱡᱟᱫ, ᱥᱚᱢᱟᱱ, ᱥᱤᱝᱜᱚᱞ ᱡᱤᱨᱚ
p(ᱫᱟᱨᱟᱭ ᱛᱮ 17)
1/37
p(150 ᱜᱷᱩᱨᱟᱹᱣ ᱨᱮ 17 ᱵᱟᱝ)
about 1.64%
ᱯᱷᱚᱭᱥᱚᱞᱟ
ᱵᱷᱨᱚᱢ

ᱫᱟᱵᱤ

ᱠᱮᱥᱤᱱᱳ ᱠᱚ ᱢᱩᱪᱟᱹᱫ 20 ᱠᱷᱚᱱ 500 ᱨᱮᱡᱟᱞᱴ ᱠᱚ ᱩᱫᱩᱜᱟ ᱟᱨ ᱟᱭᱢᱟ ᱫᱷᱟᱣ ᱡᱟᱦᱟᱸ ᱮᱞ ᱠᱚ ᱛᱤᱱᱟᱹᱜ ᱜᱟᱱ ᱚᱠᱛᱚ ᱦᱟᱹᱵᱤᱡ ᱵᱟᱝ ᱦᱮᱡ ᱟᱠᱟᱱᱟ ᱩᱱᱠᱩ “ᱴᱷᱟᱸᱰᱟ” ᱢᱮᱱᱛᱮ ᱠᱚ ᱪᱤᱱᱦᱟᱹᱣᱟ᱾ ᱠᱷᱮᱞᱚᱸᱰ ᱠᱚ ᱵᱳᱨᱰ ᱠᱚ ᱯᱟᱲᱦᱟᱣᱟ, ᱧᱮᱞ ᱠᱚ ᱧᱟᱢᱟ ᱡᱮ 17 ᱫᱚ 150 ᱜᱷᱩᱨᱟᱹᱣ ᱦᱟᱹᱵᱤᱡ ᱵᱟᱝ ᱦᱮᱡ ᱟᱠᱟᱱᱟ, ᱟᱨ ᱚᱱᱟ ᱨᱮ ᱵᱟᱡᱤ ᱠᱚ ᱞᱟᱜᱟᱣᱟ ᱪᱮᱫᱟᱜ ᱥᱮ ᱚᱱᱟ ᱦᱤᱡᱩᱜ ᱞᱟᱹᱠᱛᱤ᱾ ᱵᱤᱪᱟᱹᱨ ᱫᱚ ᱱᱚᱶᱟ ᱠᱟᱱᱟ ᱡᱮ ᱡᱮᱞᱮᱧ ᱚᱠᱛᱚ ᱨᱮ ᱡᱚᱛᱚ ᱮᱞ 1/37 ᱚᱠᱛᱚ ᱦᱮᱡᱚᱜ-ᱟ, ᱚᱱᱟᱛᱮ ᱛᱟᱭᱚᱢ ᱨᱮ ᱢᱮᱱᱟᱜ ᱮᱞ ᱫᱚ ᱥᱮᱴᱮᱨ ᱞᱟᱹᱠᱛᱤᱭᱟ᱾

ᱱᱚᱶᱟ ᱪᱮᱫᱟᱜ ᱵᱷᱩᱞ ᱠᱟᱱᱟ

In the independent uniform model, the next spin has the same probabilities whatever the recorded history. The law of large numbersᱜᱚᱱᱤᱛᱢᱟᱨᱟᱝ ᱮᱞ ᱨᱮᱭᱟᱜ ᱱᱤᱭᱚᱢᱯᱚᱨᱚᱠᱷ ᱨᱮᱭᱟᱜ ᱮᱞ ᱵᱟᱲᱦᱟᱣ ᱞᱮᱠᱷᱟᱱ ᱟᱡᱟᱫ ᱨᱮᱡᱟᱞᱴ ᱨᱮᱭᱟᱜ ᱟᱣᱥᱚᱛ ᱟᱸᱫᱟᱡᱽ ᱜᱚᱱᱚᱝ ᱴᱷᱮᱱ ᱥᱮᱴᱮᱨᱚᱜ-ᱟ᱾ᱯᱩᱨᱟᱹ ᱯᱟᱲᱦᱟᱣ → concerns proportions over growing samples; it does not require a missing number to appear more often later. A fixed past gap can become small relative to the total without being repaid.

ᱢᱤᱫ ᱮᱞ ᱠᱚᱢ ᱥᱮ ᱠᱚᱢ k ᱜᱷᱩᱨᱟᱹᱣ ᱦᱟᱹᱵᱤᱡ ᱛᱤᱱᱟᱹᱜ ᱫᱷᱟᱣ ᱵᱟᱝ ᱦᱮᱡᱚᱜ-ᱟ%
37 ᱜᱷᱩᱨᱟᱹᱣ
36.29%
74 ᱜᱷᱩᱨᱟᱹᱣ
13.17%
100 ᱜᱷᱩᱨᱟᱹᱣ
6.46%
150 ᱜᱷᱩᱨᱟᱹᱣ
1.64%
200 ᱜᱷᱩᱨᱟᱹᱣ
0.417%
%
018.1436.29%
(36/37)ᵏ᱾ ᱡᱚᱛᱚ ᱠᱮᱥᱤᱱᱳ ᱨᱮᱭᱟᱜ ᱡᱚᱛᱚ ᱪᱟᱠᱟ ᱨᱮ 37 ᱮᱞ ᱥᱟᱶ, 150 ᱜᱷᱩᱨᱟᱹᱣ ᱨᱮᱭᱟᱜ ᱵᱟᱝ ᱦᱮᱡ ᱫᱚ ᱱᱤᱛ ᱜᱮ ᱡᱟᱦᱟᱸ ᱥᱮ ᱦᱩᱭ ᱮᱫᱟ᱾

ᱱᱚᱶᱟ ᱪᱮᱫᱟᱜ ᱴᱷᱤᱠ ᱞᱮᱠᱟ ᱵᱩᱡᱷᱟᱹᱣᱚᱜ-ᱟ

ᱪᱮᱫᱟᱜ ᱥᱮ ᱡᱤᱣᱤ ᱨᱮᱭᱟᱜ ᱮᱴᱟᱜ ᱡᱚᱛᱚ ᱦᱟᱹᱴᱤᱧ ᱨᱮ, ᱵᱟᱝ ᱦᱮᱡ ᱫᱚ ᱦᱮᱡ ᱨᱮᱭᱟᱜ ᱞᱟᱦᱟ ᱠᱟᱛᱷᱟ ᱮᱢᱟ: ᱵᱟᱨ ᱜᱮᱞ ᱢᱤᱱᱤᱴ ᱦᱟᱹᱵᱤᱡ ᱵᱟᱝ ᱦᱮᱡ ᱟᱠᱟᱱ ᱵᱚᱥ ᱫᱚ ᱫᱟᱨᱟᱭ ᱛᱮ ᱢᱚᱬᱮ ᱢᱤᱱᱤᱴ ᱨᱮ ᱦᱮᱡ ᱨᱮᱭᱟᱜ ᱵᱮᱥᱤ ᱥᱚᱢᱵᱷᱟᱵᱱᱟ᱾ ᱚᱱᱟ ᱫᱚ ᱵᱚᱥ ᱫᱚ ᱚᱠᱛᱚ ᱛᱟᱹᱞᱤᱠᱟᱹ ᱞᱮᱠᱟᱛᱮ ᱪᱟᱞᱟᱜ ᱠᱟᱨᱚᱬ ᱛᱮ᱾ ᱨᱩᱞᱮᱴ ᱨᱮ ᱚᱠᱛᱚ ᱛᱟᱹᱞᱤᱠᱟᱹ ᱵᱟᱹᱱᱩᱜ-ᱟ; ᱚᱱᱰᱮ ᱢᱤᱫ ᱥᱚᱢᱵᱷᱟᱵᱱᱟᱜᱚᱱᱤᱛᱥᱚᱢᱵᱷᱟᱵᱱᱟᱢᱤᱫ ᱨᱮᱡᱟᱞᱴ ᱨᱮᱭᱟᱜ ᱥᱚᱢᱵᱷᱟᱵᱱᱟ, 0 ᱠᱷᱚᱱ 1 ᱟᱨᱵᱟᱝ 0% ᱠᱷᱚᱱ 100%᱾ ᱢᱤᱫ ᱮᱞ: 1/37 = 2.70%᱾ᱯᱩᱨᱟᱹ ᱯᱟᱲᱦᱟᱣ → ᱢᱮᱱᱟᱜ-ᱟ, ᱟᱨ ᱥᱚᱢᱵᱷᱟᱵᱱᱟ ᱫᱚ ᱢᱤᱫ ᱠᱳᱴᱟ ᱵᱟᱝ ᱠᱟᱱᱟ᱾ ᱢᱚᱱ ᱨᱮᱭᱟᱜ ᱵᱩᱡᱷᱟᱹᱣ ᱫᱚ ᱡᱟᱹᱥᱛᱤ ᱫᱩᱱᱤᱭᱟᱹ ᱞᱟᱹᱜᱤᱫ ᱴᱷᱤᱠ ᱟᱨ ᱪᱟᱠᱟ ᱞᱟᱹᱜᱤᱫ ᱵᱷᱩᱞ᱾

ᱯᱚᱨᱚᱠᱷ

Use a simulation to compare a number selected after a long absence with a number selected in advance. Under independent uniform spins, both have the same next-spin probability. Finite samples will differ, sometimes substantially.

Real-world data can also be used to investigate bias or dependence, but that requires a defined test, enough data for the effect being studied and independent validation. A “hot” or “cold” label alone establishes neither.

ᱱᱢᱩᱱᱟ ᱢᱟᱯᱨᱩᱞᱮᱴ ᱨᱮᱡᱟᱞᱴ: ᱵᱳᱨᱰ ᱯᱟᱲᱦᱟᱣ

ᱩᱞᱴᱟᱹ ᱵᱷᱨᱚᱢ

“It is hot, so it will keep winning” also lacks support under the independent uniform model. For any selected number, the probability of at least one hit in the next 37 spins is 1 − (36/37)^37, about 63.71%. That is a theoretical probability, not the guaranteed result of a short demo session.

ᱠᱩᱠᱞᱤ ᱠᱚ

ᱠᱮᱥᱤᱱᱳ ᱫᱚ ᱠᱷᱮᱞᱚᱸᱰ ᱠᱚ ᱵᱷᱩᱞᱟᱹᱣ ᱞᱟᱹᱜᱤᱫ ᱴᱷᱟᱸᱰᱟ ᱮᱞ ᱠᱚ ᱩᱫᱩᱜᱟ ᱥᱮ?

ᱠᱷᱮᱞᱚᱸᱰ ᱠᱚ ᱠᱩᱞᱤ ᱠᱟᱛᱮ ᱠᱚ ᱩᱫᱩᱜᱟ᱾ ᱵᱳᱨᱰ ᱫᱚ ᱴᱷᱤᱠ ᱟᱨ ᱵᱮᱠᱟᱹᱢᱤ᱾

500 ᱜᱷᱩᱨᱟᱹᱣ ᱦᱟᱹᱵᱤᱡ ᱵᱟᱝ ᱦᱮᱡ ᱟᱠᱟᱱ ᱮᱞ ᱫᱚ?

(36/37)⁵⁰⁰ ≈ 0.0001%: ᱵᱤᱨᱟᱹᱞ, ᱟᱨ ᱫᱟᱨᱟᱭ ᱛᱮ ᱜᱷᱩᱨᱟᱹᱣ ᱨᱮ ᱮᱱᱦᱚᱸ 1/37᱾ ᱡᱩᱫᱤ ᱢᱤᱫ ᱪᱟᱠᱟ ᱨᱮ ᱱᱚᱶᱟ ᱡᱟᱣᱨᱟ ᱦᱩᱭᱩᱜ ᱠᱷᱟᱱ, ᱚᱱᱟ ᱫᱚ ᱵᱟᱭᱟᱥ ᱨᱮᱭᱟᱜ ᱠᱩᱞᱤ ᱠᱟᱱᱟ, ᱟᱨ ᱛᱮᱞᱟ ᱫᱚ ᱚᱱᱟ ᱮᱞ ᱠᱷᱚᱱ ᱥᱟᱸᱦᱤᱡ ᱛᱟᱦᱮᱸ ᱦᱩᱭᱩᱜ-ᱟ᱾

ᱡᱟᱦᱟᱸ ᱮᱞ ᱫᱚᱨᱚᱢ ᱦᱚᱸ ᱦᱤᱡᱩᱜ ᱞᱟᱹᱠᱛᱤ ᱠᱟᱱᱟ ᱥᱮ?

ᱵᱟᱝ᱾

ᱞᱟᱸᱫᱟ ᱞᱟᱹᱜᱤᱫ ᱠᱚᱢᱤᱠ

ᱥᱚᱢᱵᱷᱟᱵᱱᱟ ᱫᱚ ᱢᱤᱫ ᱠᱳᱴᱟ ᱵᱟᱝ ᱠᱟᱱᱟ᱾

ᱥᱚᱢᱵᱷᱟᱵᱱᱟ ᱫᱚ ᱢᱤᱫ ᱠᱳᱴᱟ ᱵᱟᱝ ᱠᱟᱱᱟ᱾

ᱯᱷᱮᱰᱟᱛ

  1. Feller, W. An Introduction to Probability Theory and Its Applications, Vol. 1. Wiley, 3rd ed., 1968᱾ ᱡᱩᱣᱟᱲᱤ ᱨᱮᱭᱟᱜ ᱯᱩᱸᱡᱤ ᱢᱩᱪᱟᱹᱫ, ᱟᱫᱷᱭᱟᱭ XIV᱾
  2. Ethier, S. N. The Doctrine of Chances: Probabilistic Aspects of Gambling. Springer, 2010᱾ ᱟᱫᱷᱭᱟᱭ 13, ᱨᱩᱞᱮᱴ᱾
ᱢᱚᱰᱮᱞ: ᱭᱩᱨᱚᱯᱤᱭᱟᱱ, ᱥᱤᱝᱜᱚᱞ ᱡᱤᱨᱚᱦᱤᱥᱟᱹᱵ ᱦᱚᱨᱟ ↗ᱯᱷᱮᱰᱟᱛ ↗