ꯔꯨꯂꯦꯠꯀꯤ ꯃꯦꯊꯃꯦꯇꯤꯛꯁ · ꯋꯥꯈꯜꯂꯣꯟ · ꯃꯤꯅꯤꯠ 7

ꯃꯇꯝ ꯁꯥꯡꯅ ꯅꯛꯅꯕ: ꯔꯨꯂꯦꯠꯇ ꯑꯆꯧꯕ ꯃꯁꯤꯡꯒꯤ ꯅꯤꯌꯝ

ꯑꯦꯚꯔꯦꯖ ꯑꯁꯤ −2.70%ꯗ ꯀꯔꯝꯅ ꯅꯛꯊꯔꯀꯄꯒꯦ, ꯑꯃꯁꯨꯡ ꯃꯁꯤꯅ ꯑꯩꯈꯣꯌꯅ ꯈꯜꯂꯤꯕꯗꯒꯤ ꯀꯔꯤꯒꯤ ꯃꯇꯝ ꯍꯦꯟꯅ ꯆꯪꯕꯒꯦ꯫

ꯊꯨꯅ ꯄꯥꯎꯈꯨꯝ

ꯁ꯭ꯄꯤꯟ ꯍꯦꯟꯒꯠꯂꯛꯄꯒꯥ ꯂꯣꯌꯅꯅ ꯁ꯭ꯄꯤꯟ ꯈꯨꯗꯤꯡꯗ ꯃꯥꯡꯕ ꯑꯦꯚꯔꯦꯖ ꯁꯦꯜ ꯑꯁꯤ −1/37ꯗ ꯅꯛꯊꯔꯀꯏ, ꯑꯃꯁꯨꯡ ꯃꯁꯤꯒꯤ ꯁ꯭ꯞꯔꯦꯗ ꯑꯁꯤ 1/√N ꯒꯨꯝꯅ ꯍꯟꯊꯔꯛꯂꯤ꯫ ꯑꯉꯥꯡꯕꯗ ꯁ꯭ꯄꯤꯟ 1,000 ꯇꯥꯔꯕꯥ ꯃꯇꯨꯡꯗ ꯑꯦꯚꯔꯦꯖ ꯑꯁꯤ ꯌꯥꯝꯅꯥ ꯀꯌꯥꯗ ꯃꯁꯤꯒꯤ ꯐꯪꯒꯗꯕ ꯃꯃꯂꯗꯒꯤ ±3%ꯒꯤ ꯃꯅꯨꯡꯗ ꯂꯩ꯫

ꯁ꯭ꯄꯤꯟ ꯈꯨꯗꯤꯡꯗ ꯐꯪꯒꯗꯕ
−2.70%
N ꯃꯇꯨꯡꯒꯤ ꯁ꯭ꯞꯔꯦꯗ
σ/√N
ꯑꯉꯥꯡꯕ, ꯑꯦꯚꯔꯦꯖꯒꯤ SD ≈ 1%
ꯁ꯭ꯄꯤꯟ 10,000
ꯄꯨꯝꯅꯃꯛ ꯁꯦꯜ ꯅꯛꯊꯔꯀꯗꯦ
ꯂꯥꯞꯊꯣꯛꯂꯛꯏ

ꯑꯆꯧꯕ ꯃꯁꯤꯡꯒꯤ ꯅꯤꯌꯝ

For independent, identically distributed bets with finite expectation, the running average tends toward the expected valueꯃꯦꯊꯃꯦꯇꯤꯛꯁꯐꯪꯒꯗꯕ ꯃꯃꯜꯃꯇꯝ ꯁꯥꯡꯅ ꯌꯦꯡꯂꯕꯗ ꯕꯦꯠ ꯑꯃꯒꯤ ꯑꯦꯚꯔꯦꯖ ꯃꯍꯩ꯫ ꯃꯔꯨꯑꯣꯏꯕ ꯌꯨꯔꯣꯄꯤꯌꯟ ꯕꯦꯠ ꯈꯨꯗꯤꯡꯗ ꯊꯝꯂꯕ ꯁꯦꯜꯒꯤ −1/37꯫ꯃꯄꯨꯡ ꯄꯥꯕꯤꯌꯨ →. After N standard single-zeroꯆꯛꯔꯖꯤꯔꯣꯆꯞ ꯃꯥꯟꯅꯕ, ꯗꯖꯟ ꯅꯠꯔꯒ ꯀꯂꯝ ꯕꯦꯠ ꯑꯃꯠꯇꯅ ꯀꯨꯞꯇꯕ ꯑꯁꯪꯕ ꯄꯣꯀꯦꯠ ꯑꯃꯈꯛꯇ; ꯍꯥꯎꯁ ꯑꯦꯖꯀꯤ ꯍꯧꯔꯀꯐꯝꯅꯤ꯫ ꯑꯇꯣꯞꯄ ꯃꯃꯤꯡ: ꯁꯤꯡꯒꯜ ꯖꯤꯔꯣ, 0꯫ꯃꯄꯨꯡ ꯄꯥꯕꯤꯌꯨ → even-money betsꯕꯦꯠꯆꯞ ꯃꯥꯟꯅꯕ ꯕꯦꯠꯑꯉꯥꯡꯕ/ꯑꯃꯨꯕ, ꯕꯤꯖꯣꯔ/ꯖꯣꯔ, ꯑꯋꯥꯡꯕ/ꯑꯅꯦꯝꯕ꯫ ꯃꯁꯤꯡ 18, 1:1 ꯄꯦꯑꯥꯎꯠ꯫ꯃꯄꯨꯡ ꯄꯥꯕꯤꯌꯨ → at one unitꯁꯦꯜꯌꯨꯅꯤꯠꯁꯤꯁ꯭ꯇꯦꯝ ꯅꯠꯔꯒ ꯁꯥꯟꯅꯔꯤꯕ ꯑꯃꯅ ꯁꯤꯖꯤꯟꯅꯕ ꯃꯔꯨꯑꯣꯏꯕ ꯊꯝꯂꯕ ꯁꯦꯜ꯫ ꯄ꯭ꯔꯣꯒ꯭ꯔꯦꯁꯅꯁꯤꯡ ꯌꯨꯅꯤꯠꯇ ꯁꯟꯗꯣꯛꯂꯤ, ꯃꯔꯝ ꯑꯗꯨꯅ ꯃꯈꯣꯌ ꯀꯔꯤꯒꯨꯝꯕ ꯇꯦꯕꯜꯗ ꯆꯠꯂꯤ꯫ꯃꯄꯨꯡ ꯄꯥꯕꯤꯌꯨ →, its expectation is −1/37 and its standard deviationꯃꯦꯊꯃꯦꯇꯤꯛꯁꯁ꯭ꯇꯦꯟꯗꯔꯗ ꯗꯤꯚꯤꯌꯦꯁꯟꯚꯦꯌꯔꯤꯌꯦꯟꯁꯀꯤ ꯁ꯭ꯀ꯭ꯋꯦꯌꯔ ꯔꯨꯠ; ꯃꯍꯩ ꯑꯃꯅ ꯐꯪꯒꯗꯕ ꯃꯃꯜꯗꯒꯤ ꯌꯥꯝꯅ ꯂꯥꯞꯄ ꯃꯑꯣꯡ꯫ꯃꯄꯨꯡ ꯄꯥꯕꯤꯌꯨ → is about 1/√N. That spread is about 10% of the stake after 100 spins, 3.16% after 1,000 and 1% after 10,000. More observations make the edge easier to distinguish from random variation; no fixed spin count guarantees a particular observation.

ꯁ꯭ꯄꯤꯟꯁ꯭ꯄꯤꯟ ꯈꯨꯗꯤꯡꯒꯤ ꯑꯦꯚꯔꯦꯖꯀꯤ σ (ꯆꯞ ꯃꯥꯟꯅꯕ)σ ꯒꯤ ꯁꯔꯨꯛ ꯑꯣꯏꯅ ꯍꯥꯎꯁ ꯑꯦꯖ
1031.6%0.09
10010.0%0.27
1,0003.2%0.85
10,0001.0%2.70
100,0000.3%8.54
ꯊꯝꯂꯕ ꯌꯨꯅꯤꯠ ꯑꯃꯒꯤꯗꯃꯛ꯫ ꯍꯥꯎꯁ ꯑꯦꯖ ꯑꯁꯤ ꯁ꯭ꯄꯤꯟ 1,000 ꯑꯃꯁꯨꯡ 10,000 ꯒꯤ ꯃꯔꯛꯇ ꯁ꯭ꯇꯦꯇꯤꯁ꯭ꯇꯤꯀꯦꯜ ꯑꯣꯏꯅ ꯃꯌꯦꯛ ꯁꯦꯡꯂꯀꯏ꯫

ꯀꯔꯤ ꯅꯛꯊꯔꯀꯄꯒꯦ ꯑꯃꯁꯨꯡ ꯀꯔꯤ ꯅꯛꯊꯔꯀꯗꯕꯒꯦ

ꯁ꯭ꯄꯤꯟ ꯈꯨꯗꯤꯡꯒꯤ ꯑꯦꯚꯔꯦꯖ ꯑꯁꯤ ꯅꯛꯊꯔꯀꯏ꯫ ꯄꯨꯝꯅꯃꯛ ꯁꯦꯜ ꯑꯗꯨ ꯅꯛꯊꯔꯀꯗꯦ: N ꯍꯦꯟꯒꯠꯂꯛꯄꯒꯥ ꯂꯣꯌꯅꯅ ꯃꯁꯤ ꯖꯤꯔꯣꯗꯒꯤ ꯍꯦꯟꯅ ꯂꯥꯞꯊꯣꯛꯂꯛꯏ, ꯐꯪꯒꯗꯕ ꯃꯑꯣꯡꯗ N ꯒꯤ 2.7% ꯃꯔꯥꯡ ꯀꯥꯏꯅ, ꯃꯁꯤꯒꯤ ꯃꯅꯥꯛꯇ √N ꯒꯤ ꯁ꯭ꯞꯔꯦꯗ ꯂꯩꯅꯥ꯫ $10 ꯗ ꯁ꯭ꯄꯤꯟ 10,000 ꯁꯥꯟꯈꯤꯕ ꯁꯥꯟꯅꯔꯤꯕ ꯑꯃꯥ $2,700 ꯃꯥꯡꯒꯅꯤ ꯍꯥꯌꯅ ꯊꯥꯖꯩ, ꯃꯁꯤꯗꯒꯤ $1,000 ꯍꯦꯟꯕ ꯅꯠꯔꯒ ꯍꯟꯕ꯫ ꯑꯉꯥꯡꯕꯕꯦꯠꯑꯉꯥꯡꯕꯑꯉꯥꯡꯕ ꯃꯁꯤꯡ 18 ꯗ ꯊꯝꯕ ꯆꯞ ꯃꯥꯟꯅꯕ ꯕꯦꯠ꯫ ꯐ꯭ꯔꯦꯟꯆ: ꯔꯨꯖ꯫ꯃꯄꯨꯡ ꯄꯥꯕꯤꯌꯨ → ꯊꯣꯛꯄꯒꯤ ꯆꯠꯂꯤꯕ ꯁꯔꯨꯛ ꯑꯁꯤ ꯌꯥꯝꯅꯥ ꯃꯃꯥꯡꯗꯒꯤ 48.65% ꯒꯤ ꯃꯅꯥꯛꯇ ꯂꯦꯞꯂꯦ; ꯆꯠꯂꯤꯕ ꯕꯦꯂꯦꯟꯁ ꯑꯗꯨꯗꯤ ꯀꯗꯥꯏꯗꯁꯨ ꯂꯦꯞꯂꯃꯗꯦ꯫

ꯁꯥꯡꯕ ꯁꯦꯁꯟ ꯇꯔꯥ: ꯑꯉꯥꯡꯕꯗ $10 ꯒꯤ ꯁ꯭ꯄꯤꯟ 1,000ꯍꯧꯔꯛꯐꯃꯒꯥ ꯆꯥꯡꯗꯝꯅꯕ ꯃꯍꯩ
+$440$0−$660+$160−$3300125250375500ꯁ꯭ꯄꯤꯟꯃꯍꯩ
ꯁꯦꯁꯟ 1,000 × ꯁ꯭ꯄꯤꯟ 500 · ꯃꯆꯨ ꯑꯃꯨ-ꯂꯩꯃꯜ p10/p25/p75/p90, ꯑꯉꯧ-ꯑꯁꯪꯕ ꯃꯤꯗꯤꯌꯥꯟ, ꯃꯌꯥꯏꯕ ꯈꯨꯗꯝ 79 · ꯁ꯭ꯄꯤꯟ ꯑꯃꯗ ꯀ꯭ꯕꯥꯟꯇꯥꯏꯜ ꯎꯅꯕ ꯆꯥꯔꯠꯀꯤ ꯃꯊꯛꯇ ꯃꯧꯁ ꯆꯠꯍꯅꯕꯤꯌꯨ
  • p10 −$440
  • p25 −$280
  • ꯃꯤꯗꯤꯌꯥꯟ −$140
  • p75 +$0
  • p90 +$140
ꯁꯦꯁꯟ 1,000 × ꯁ꯭ꯄꯤꯟ 500 · y-ꯑꯦꯛꯁꯤꯁ 1ꯁꯨꯕ–99ꯁꯨꯕ ꯄꯔꯁꯦꯟꯇꯥꯏꯜ ꯐꯥꯑꯣꯕꯥ ꯀꯛꯊꯠꯂꯤ (ꯑꯔꯣꯏꯕ ꯃꯁꯤꯡ −$940ꯗꯒꯤ +$660ꯐꯥꯑꯣꯕꯥ) · ꯁ꯭ꯄꯤꯟ 500 · ꯍꯧꯔꯛꯐꯝ $10 · ꯆꯞ ꯃꯥꯟꯅꯕ
ꯁ꯭ꯄꯤꯟ 1,000 ꯌꯧꯕꯥ ꯃꯇꯃꯗ ꯂꯥꯏꯟ ꯀꯌꯥꯅ ꯖꯤꯔꯣꯒꯤ ꯃꯈꯥꯗ ꯂꯩꯔꯦ ꯑꯃꯁꯨꯡ ꯁ꯭ꯞꯔꯦꯗ ꯑꯁꯤ ꯍꯧꯖꯤꯛ ꯐꯥꯑꯣꯕ ꯄꯥꯛꯂꯤ꯫ ꯍꯥꯎꯁ ꯑꯦꯖꯅ ꯃꯥꯏ ꯄꯥꯛꯂꯤ; ꯚꯦꯌꯔꯤꯌꯦꯟꯁꯁꯨ ꯍꯧꯖꯤꯛ ꯐꯥꯑꯣꯕ ꯈꯣꯟꯖꯦꯜ ꯆꯥꯎꯅ ꯂꯩꯔꯤ꯫

ꯈꯜꯂꯕ ꯑꯗꯨ ꯀꯔꯤꯒꯤ ꯑꯔꯥꯟꯕꯒꯦ

2.70% ꯍꯥꯌꯕꯁꯤ ꯌꯥꯝꯅꯥ ꯆꯥꯎꯕ ꯑꯣꯏꯅ ꯇꯥꯕꯅꯥ ꯃꯤꯁꯤꯡꯅ ꯍꯥꯎꯁ ꯑꯦꯖ ꯑꯁꯤ ꯌꯥꯝꯅꯥ ꯊꯨꯅ ꯎꯕ ꯐꯪꯒꯅꯤ ꯍꯥꯌꯅ ꯈꯜꯂꯤ, ꯑꯃꯁꯨꯡ ꯑꯆꯧꯕ ꯃꯁꯤꯡꯒꯤ ꯅꯤꯌꯝꯅꯃꯦꯊꯃꯦꯇꯤꯛꯁꯑꯆꯧꯕ ꯃꯁꯤꯡꯒꯤ ꯅꯤꯌꯝꯍꯣꯠꯅꯕꯒꯤ ꯃꯁꯤꯡ ꯍꯦꯅꯒꯠꯂꯀꯄꯒ ꯂꯣꯌꯅꯅ ꯃꯔꯤ ꯂꯩꯅꯗꯕ ꯃꯍꯩꯁꯤꯡꯒꯤ ꯑꯦꯚꯔꯦꯖ ꯑꯗꯨ ꯐꯪꯒꯗꯕ ꯃꯃꯂꯗ ꯅꯛꯂꯛꯏ꯫ꯃꯄꯨꯡ ꯄꯥꯕꯤꯌꯨ → ꯑꯦꯚꯔꯦꯖꯁꯤꯡ ꯅꯛꯊꯔꯀꯏ ꯍꯥꯌꯕꯅꯥ ꯁꯥꯡꯕ ꯁ꯭ꯇ꯭ꯔꯤꯛꯁꯤꯡ “ꯆꯞ ꯃꯥꯟꯅꯍꯟꯒꯅꯤ” ꯍꯥꯌꯅ ꯈꯜꯂꯤ꯫ ꯈꯜꯂꯕ ꯑꯅꯤ ꯑꯁꯤꯅ ꯆꯥꯡ ꯑꯗꯨ ꯑꯔꯥꯟꯅ ꯂꯧꯏ: ꯍꯥꯎꯁ ꯑꯦꯖ ꯑꯁꯤ ꯁ꯭ꯄꯤꯟ ꯑꯃꯒꯤ ꯚꯦꯌꯔꯤꯌꯦꯟꯁꯀꯥꯃꯦꯊꯃꯦꯇꯤꯛꯁꯚꯦꯌꯔꯤꯌꯦꯟꯁꯃꯍꯩꯁꯤꯡ ꯐꯪꯒꯗꯕ ꯃꯃꯂꯒꯤ ꯑꯀꯣꯌꯕꯗ ꯀꯌꯥ ꯁꯟꯗꯣꯛꯅ ꯂꯩꯕꯒꯦ ꯍꯥꯌꯕꯒꯤ ꯆꯥꯡ꯫ ꯁ꯭ꯇ꯭ꯔꯦꯠ-ꯑꯞꯇ ꯋꯥꯡꯏ, ꯆꯞ ꯃꯥꯟꯅꯕ ꯕꯦꯠꯁꯤꯡꯗ ꯅꯦꯝꯏ꯫ꯃꯄꯨꯡ ꯄꯥꯕꯤꯌꯨ → ꯆꯥꯡꯗꯝꯅꯕꯗ ꯑꯄꯤꯛꯄꯅꯤ (σ = ꯌꯨꯅꯤꯠ 1, EV = 0.027), ꯃꯔꯝ ꯑꯗꯨꯅ ꯃꯁꯤ ꯁꯤꯒ꯭ꯃꯥ ꯑꯃꯗ ꯎꯕ ꯐꯪꯅꯕ (1/0.027)² ≈ 1,400 ꯁ꯭ꯄꯤꯟ ꯃꯊꯧ ꯇꯥꯏ꯫ ꯑꯃꯁꯨꯡ ꯑꯦꯚꯔꯦꯖ ꯅꯛꯊꯔꯛꯄ ꯑꯁꯤꯅ ꯄꯨꯝꯅꯃꯛ ꯁꯦꯜ ꯑꯗꯨ ꯍꯜꯂꯛꯍꯟꯗꯦ; ꯃꯁꯤꯅ ꯃꯃꯥꯡꯒꯤ ꯍꯣꯡꯂꯛ-ꯍꯣꯡꯊꯛꯁꯤꯡ ꯑꯗꯨ ꯌꯥꯎꯊꯣꯛꯍꯜꯂꯤ ꯈꯛꯇꯅꯤ꯫

ꯑꯇꯣꯞꯄ ꯕꯦꯠꯁꯤꯡꯒꯤ ꯅꯛꯊꯔꯛꯄ

For straight-up bets, σ is about 5.84 per unit, compared with about 1 for even-money bets. Using exact values, expected lossꯃꯦꯊꯃꯦꯇꯤꯛꯁꯃꯥꯡꯒꯗꯕ ꯁꯦꯜꯐꯪꯒꯗꯕ ꯃꯃꯂꯕꯨ ꯕꯦꯠ ꯊꯝꯂꯕ ꯁꯦꯜ ꯄꯨꯝꯅꯃꯛꯅ ꯃꯅꯥꯡꯕ; ꯁꯦꯁꯟ ꯅꯠꯔꯒ ꯃꯇꯝ ꯑꯃꯒꯤ ꯃꯅꯨꯡꯗ ꯃꯥꯡꯕ ꯑꯦꯚꯔꯦꯖ ꯁꯦꯜ꯫ꯃꯄꯨꯡ ꯄꯥꯕꯤꯌꯨ → equals one standard deviation of the total at 46,656 straight-up spins and 1,368 even-money spins. This compares scales of variation; one standard deviation is not a threshold that establishes an edge.

ꯁꯤꯁ꯭ꯇꯦꯝ ꯆꯥꯡꯌꯦꯡ ꯇꯧꯕꯗ ꯃꯁꯤꯒꯤ ꯃꯔꯣꯜ

A positive sample average does not establish that a system has positive expected value. A useful test specifies the rules, stakes, stopping rule and success criterion before examining results, and accounts for trying multiple systems. Under fixed-stake single-zero red betting, a profit after exactly 100 spins has probabilityꯃꯦꯊꯃꯦꯇꯤꯛꯁꯊꯣꯛꯄꯒꯤ ꯈꯨꯗꯣꯡꯆꯥꯕꯃꯍꯩ ꯑꯃꯥ ꯊꯣꯛꯄꯒꯤ ꯈꯨꯗꯣꯡꯆꯥꯕ, 0 ꯗꯒꯤ 1 ꯐꯥꯑꯣꯕ ꯅꯠꯔꯒ 0% ꯗꯒꯤ 100% ꯐꯥꯑꯣꯕ꯫ ꯃꯁꯤꯡ ꯑꯃꯥ: 1/37 = 2.70%꯫ꯃꯄꯨꯡ ꯄꯥꯕꯤꯌꯨ → about 35.53%. Simulation can test a stated model, but cannot prove that a real wheel behaves like that model.

ꯐꯣꯔꯃꯨꯂꯥ

ꯁ꯭ꯄꯤꯟ ꯈꯨꯗꯤꯡꯒꯤ ꯑꯦꯚꯔꯦꯖ → −1/37
σ(ꯑꯦꯚꯔꯦꯖ) = σ / √N
E[ꯄꯨꯝꯅꯃꯛ] = −N/37 · σ(ꯄꯨꯝꯅꯃꯛ) = σ√N

ꯋꯥꯍꯪꯁꯤꯡ

ꯑꯩꯅ ꯃꯇꯝ ꯁꯥꯡꯅ ꯁꯥꯟꯂꯕꯗꯤ ꯑꯔꯣꯏꯕꯗ ꯀꯥꯟꯅꯕ ꯐꯪꯂꯣꯏꯗ꯭ꯔꯥ?

Playing longer does not guarantee that you will get ahead. In the idealized fixed one-unit red-bet model with no bankrollꯁꯦꯜꯕꯦꯡꯀꯔꯣꯜꯁꯥꯟꯅꯅꯕ ꯇꯣꯉꯥꯟꯅ ꯊꯝꯂꯕ ꯁꯦꯜ ꯄꯨꯝꯅꯃꯛ꯫ ꯕꯦꯡꯀꯔꯣꯜ ꯃꯦꯅꯦꯖꯃꯦꯟꯠ ꯍꯥꯌꯕꯁꯤ ꯃꯗꯨꯒꯤ ꯀꯌꯥ ꯕꯦꯠ ꯈꯨꯗꯤꯡꯗ ꯑꯃꯁꯨꯡ ꯁꯦꯁꯟ ꯈꯨꯗꯤꯡꯗ ꯈꯨꯗꯣꯡꯊꯤꯕꯗ ꯊꯃꯒꯗꯒꯦ ꯍꯥꯌꯕ ꯋꯥꯔꯦꯞ ꯂꯧꯕꯅꯤ꯫ꯃꯄꯨꯡ ꯄꯥꯕꯤꯌꯨ → limit, the chance of ever reaching a one-unit profit from zero is 18/19, about 94.7%. A finite bankroll lowers that chance. Continuing indefinitely with negative expectation eventually drives the total downward under this model.

ꯔꯨꯂꯦꯠꯇ ꯑꯆꯧꯕ ꯃꯁꯤꯡꯒꯤ ꯅꯤꯌꯝꯅ ꯆꯛꯔ ꯑꯁꯤꯅ ꯃꯁꯥꯃꯛ ꯆꯨꯃꯊꯣꯛꯏ ꯍꯥꯌꯕꯔꯥ?

ꯅꯠꯇꯦ꯫ ꯃꯁꯤꯒꯤ ꯑꯔꯊꯗꯤ ꯃꯃꯥꯡꯒꯤ ꯍꯣꯡꯂꯛ-ꯍꯣꯡꯊꯛꯁꯤꯡ ꯑꯁꯤ ꯍꯦꯟꯅ ꯆꯥꯎꯕ ꯁꯦꯝꯄꯜ ꯑꯃꯒꯤ ꯑꯄꯤꯛꯄ ꯁꯔꯨꯛ ꯑꯃꯥ ꯑꯣꯏꯔꯛꯏ꯫

ꯃꯍꯩ ꯑꯃꯥ ꯊꯥꯖꯕ ꯌꯥꯅꯕ ꯁ꯭ꯄꯤꯟ ꯀꯌꯥ ꯃꯊꯧ ꯇꯥꯒꯅꯤ?

ꯆꯞ ꯃꯥꯟꯅꯕ ꯕꯦꯠꯀꯤꯗꯃꯛ, ꯍꯥꯎꯁ ꯑꯦꯖ ꯃꯌꯦꯛ ꯁꯦꯡꯅ ꯎꯕ ꯐꯪꯅꯕ 10,000 ꯃꯔꯥꯡ ꯀꯥꯏꯅ꯫ ꯁ꯭ꯇ꯭ꯔꯦꯠ-ꯑꯞꯀꯤꯗꯃꯛ, 50,000 ꯃꯔꯥꯡ ꯀꯥꯏꯅ꯫

ꯃꯦꯊꯃꯦꯇꯤꯛꯁꯀꯥ ꯁꯥꯟꯅꯕ

ꯃꯁꯤꯁꯤꯡ ꯑꯍꯥꯅꯕꯗ ꯅꯍꯥꯛ ꯃꯁꯥꯃꯛꯅ ꯍꯣꯠꯅꯕꯤꯌꯨ꯫ ꯄꯥꯎꯈꯨꯝ ꯈꯨꯗꯤꯡꯃꯛ ꯊꯨꯡꯗꯨꯅ ꯊꯝꯂꯤ; ꯁꯦꯝ ꯁꯥꯔꯕ ꯃꯇꯃꯗ ꯍꯥꯡꯗꯣꯛꯄꯤꯌꯨ꯫

  1. ꯃꯔꯨꯞ ꯑꯅꯤꯅ ꯆꯨꯝꯃꯤꯟꯅꯕ ꯁꯥꯟꯅꯕꯃꯦꯊꯃꯦꯇꯤꯛꯁꯆꯨꯝꯃꯤꯟꯅꯕ ꯁꯥꯟꯅꯕꯀꯅꯥꯒꯨꯝꯕ ꯃꯥꯏꯀꯩ ꯑꯃꯠꯇꯗ ꯑꯦꯖ ꯂꯩꯇꯕ ꯁꯥꯟꯅꯕ: ꯕꯦꯠ ꯈꯨꯗꯤꯡꯃꯛꯀꯤ ꯐꯪꯒꯗꯕ ꯃꯃꯜ ꯖꯤꯔꯣꯅꯤ꯫ꯃꯄꯨꯡ ꯄꯥꯕꯤꯌꯨ → ꯑꯃꯒꯤ ꯔꯥꯎꯟꯗ 20 ꯁꯥꯟꯅꯩ, ꯀꯣꯏꯟ ꯊꯥꯗꯣꯛꯄ ꯅꯠꯔꯒꯥ ꯖꯤꯔꯣꯁꯤꯡꯆꯛꯔꯖꯤꯔꯣꯆꯞ ꯃꯥꯟꯅꯕ, ꯗꯖꯟ ꯅꯠꯔꯒ ꯀꯂꯝ ꯕꯦꯠ ꯑꯃꯠꯇꯅ ꯀꯨꯞꯇꯕ ꯑꯁꯪꯕ ꯄꯣꯀꯦꯠ ꯑꯃꯈꯛꯇ; ꯍꯥꯎꯁ ꯑꯦꯖꯀꯤ ꯍꯧꯔꯀꯐꯝꯅꯤ꯫ ꯑꯇꯣꯞꯄ ꯃꯃꯤꯡ: ꯁꯤꯡꯒꯜ ꯖꯤꯔꯣ, 0꯫ꯃꯄꯨꯡ ꯄꯥꯕꯤꯌꯨ → ꯍꯜꯂꯛꯄ ꯑꯉꯥꯡꯕꯕꯦꯠꯑꯉꯥꯡꯕꯑꯉꯥꯡꯕ ꯃꯁꯤꯡ 18 ꯗ ꯊꯝꯕ ꯆꯞ ꯃꯥꯟꯅꯕ ꯕꯦꯠ꯫ ꯐ꯭ꯔꯦꯟꯆ: ꯔꯨꯖ꯫ꯃꯄꯨꯡ ꯄꯥꯕꯤꯌꯨ → ꯑꯃꯁꯨꯡ ꯑꯃꯨꯕꯕꯦꯠꯑꯃꯨꯕꯑꯃꯨꯕ ꯃꯁꯤꯡ 18 ꯗ ꯊꯝꯕ ꯆꯞ ꯃꯥꯟꯅꯕ ꯕꯦꯠ꯫ ꯐ꯭ꯔꯦꯟꯆ: ꯅꯣꯌꯥꯔ꯫ꯃꯄꯨꯡ ꯄꯥꯕꯤꯌꯨ →, ꯑꯃꯁꯨꯡ ꯆꯠꯂꯤꯕ ꯁ꯭ꯀꯣꯔ ꯊꯝꯃꯤ꯫ ꯃꯤ ꯈꯨꯗꯤꯡ ꯃꯥꯡꯗ ꯂꯩꯔꯤꯕ ꯔꯥꯎꯟꯗꯁꯤꯡ ꯃꯁꯤꯡ ꯊꯤꯕꯤꯌꯨ (ꯃꯥꯟꯅꯕ ꯑꯃꯥ ꯃꯃꯥꯡꯗ ꯃꯥꯡꯗ ꯂꯩꯔꯝꯕꯒꯤꯗꯃꯛ ꯊꯤꯏ)꯫ ꯀꯅꯥꯒꯤ ꯈꯨꯗꯣꯡꯆꯥꯕ ꯍꯦꯟꯅ ꯂꯩ: ꯃꯥꯡꯗ ꯂꯩꯕ ꯑꯗꯨ ꯔꯥꯎꯟꯗ 10 ꯗꯒꯤ 10 ꯌꯦꯟꯅꯕ, ꯅꯠꯔꯒꯥ ꯃꯔꯨꯞ ꯑꯃꯅ 20 ꯄꯨꯝꯅꯃꯛꯇ ꯃꯥꯡꯗ ꯂꯩꯕ?

    ꯄꯥꯎꯈꯨꯝ ꯎꯠꯄ

    ꯄꯥꯎꯈꯨꯝ ꯃꯔꯨꯞ ꯑꯃꯅ ꯔꯥꯎꯟꯗ 20 ꯄꯨꯝꯅꯃꯛꯇ ꯃꯥꯡꯗ ꯂꯩꯕꯒꯤ ꯈꯨꯗꯣꯡꯆꯥꯕ ꯌꯥꯝꯅꯥ ꯍꯦꯟꯅ ꯂꯩ: 35.2%, ꯀꯅꯥꯒꯨꯝꯕ ꯃꯔꯨꯞ ꯑꯃꯥ ꯊꯤꯔꯕꯗ, ꯆꯞ ꯃꯥꯟꯅꯕ 10 ꯗꯒꯤ 10 ꯒꯤ 6.1% ꯒꯥ ꯆꯥꯡꯗꯝꯂꯕꯗ꯫

    ꯆꯨꯝꯃꯤꯟꯅꯕ ꯆꯠꯂꯤꯕ ꯁ꯭ꯀꯣꯔ ꯑꯃꯥ ꯔꯥꯎꯟꯗ 2k ꯒꯤ ꯃꯇꯨꯡꯗ ꯆꯞ ꯃꯥꯟꯅꯕꯒꯤ ꯈꯨꯗꯣꯡꯆꯥꯕ u = C(2k, k)/2²ᵏ: ꯔꯥꯎꯟꯗ 10 ꯒꯤ ꯃꯇꯨꯡꯗ 0.246 ꯑꯃꯁꯨꯡ 20 ꯒꯤ ꯃꯇꯨꯡꯗ 0.176꯫ ꯀ꯭ꯂꯥꯁꯤꯛ ꯃꯍꯩ ꯑꯃꯥ, ꯑꯥꯔꯛ-ꯁꯥꯏꯟ ꯅꯤꯌꯝ, ꯅ ꯍꯥꯌꯔꯤ ꯃꯃꯤꯡ ꯊꯣꯅꯈꯤꯕ ꯃꯔꯨꯞ ꯑꯃꯅ ꯔꯥꯎꯟꯗ 2n ꯒꯤ ꯃꯅꯨꯡꯗ ꯆꯨꯝꯅ 2k ꯃꯥꯡꯗ ꯂꯩꯕꯒꯤ ꯈꯨꯗꯣꯡꯆꯥꯕ u(2k) × u(2n − 2k)꯫

    20 ꯄꯨꯝꯅꯃꯛꯇ ꯃꯥꯡꯗ ꯂꯩꯕ: ꯃꯔꯨꯞ ꯈꯨꯗꯤꯡꯒꯤꯗꯃꯛ u(20) × u(0) = 0.176, ꯀꯅꯥꯒꯨꯝꯕ ꯑꯃꯒꯤꯗꯃꯛ 35.2%꯫ ꯆꯞ ꯃꯥꯟꯅꯕ ꯌꯦꯟꯅꯕ: u(10) × u(10) = 0.246² = 6.1%, ꯈ꯭ꯕꯥꯏꯗꯒꯤ ꯈꯨꯗꯣꯡꯆꯥꯕ ꯍꯟꯊꯕ ꯌꯦꯟꯅꯕ꯫

    ꯃꯔꯝ ꯑꯗꯨꯅ ꯆꯨꯝꯃꯤꯟꯅꯕ ꯁꯥꯟꯅꯕ ꯑꯃꯗ ꯃꯥꯡꯗ ꯂꯩꯕ ꯑꯗꯨ ꯌꯥꯝꯅꯥ ꯇꯞꯅ ꯍꯣꯡꯏ꯫ ꯅꯨꯡꯊꯤꯜ ꯄꯨꯝꯅꯃꯛ ꯃꯥꯡꯗ ꯂꯩꯔꯝꯕ, ꯅꯠꯔꯒꯥ ꯃꯇꯨꯡꯗ ꯂꯩꯔꯝꯕ ꯁꯥꯟꯅꯔꯤꯕ ꯑꯃꯅ ꯑꯉꯛꯄ ꯀꯔꯤꯁꯨ ꯎꯈꯤꯗꯦ, ꯑꯃꯁꯨꯡ ꯃꯁꯤꯅ ꯆꯛꯔ ꯅꯠꯔꯒꯥ ꯃꯁꯛ ꯂꯧꯁꯤꯡꯒꯤ ꯃꯔꯃꯗ ꯀꯔꯤꯁꯨ ꯍꯥꯌꯗꯦ꯫ Epstein (2013) ꯅ ꯍꯦꯟꯅ ꯁꯥꯡꯕ ꯁꯥꯟꯅꯕꯁꯤꯡꯒꯤꯗꯃꯛ ꯅꯤꯌꯝ ꯑꯗꯨꯃꯛ ꯄꯤꯔꯤ: ꯃꯇꯝ ꯌꯥꯝꯈꯥꯏꯗ ꯂꯥꯛ ꯐꯕ ꯃꯥꯏꯀꯩ ꯑꯗꯨ ꯁꯥꯟꯅꯕꯒꯤ 85% ꯗꯒꯤ ꯍꯦꯟꯅ ꯃꯥꯡꯗ ꯂꯩ꯫

ꯅꯣꯛꯅꯕꯒꯤ ꯀꯃꯤꯛ

ꯑꯦꯚꯔꯦꯖ ꯑꯁꯤ ꯅꯛꯊꯔꯀꯏ꯫ ꯄꯨꯝꯅꯃꯛ ꯁꯦꯜ ꯑꯗꯨꯗꯤ ꯂꯥꯞꯊꯣꯛꯂꯛꯏ ꯈꯛꯇꯅꯤ꯫

ꯑꯦꯚꯔꯦꯖ ꯑꯁꯤ ꯅꯛꯊꯔꯀꯏ꯫ ꯄꯨꯝꯅꯃꯛ ꯁꯦꯜ ꯑꯗꯨꯗꯤ ꯂꯥꯞꯊꯣꯛꯂꯛꯏ ꯈꯛꯇꯅꯤ꯫

ꯍꯧꯔꯛꯐꯃꯁꯤꯡ

  1. Bernoulli, J. Ars Conjectandi. 1713꯫ ꯑꯆꯧꯕ ꯃꯁꯤꯡꯒꯤ ꯅꯤꯌꯝ꯫
  2. Feller, W. An Introduction to Probability Theory and Its Applications, Vol. 1. Wiley, ꯑꯦꯗꯤꯁꯟ 3, 1968꯫ ꯖꯨꯌꯥ ꯁꯥꯟꯅꯔꯤꯕꯒꯤ ꯃꯥꯡꯍꯟ ꯇꯥꯀꯍꯟꯕ, ꯆꯦꯞꯇꯔ XIV꯫
ꯃꯣꯗꯦꯜ: ꯌꯨꯔꯣꯄꯤꯌꯟ, ꯁꯤꯡꯒꯜ ꯖꯤꯔꯣꯃꯦꯊꯗꯣꯂꯣꯖꯤ ↗ꯍꯧꯔꯛꯐꯃꯁꯤꯡ ↗