ВИ НЕПРАВИЛЬНО ГРАЛИ У ГРУ ЖИТТЯ [VERITASIUM]

Scientific concepts / discoveries / phenomena

• Normal distribution (Gaussian statistics)

  • Many independent random factors add up → outcomes cluster around a mean.
  • Extreme values are extremely rare.
  • Used as a contrast case: in “normal-distribution worlds,” averages stabilize and scale is predictable.

• Power laws in nature

  • Observed when extremely large events are far more probable than a normal distribution would predict.
  • Heavy tails strongly distort averages and make prediction difficult (no meaningful finite “width” like a standard deviation).
  • Key idea: large and small events coexist across many scales.

• Pareto’s discovery of the Pareto principle / Pareto distribution

  • Vilfredo Pareto (late 1800s) analyzed income data using tax returns across European countries.
  • He found income follows an abnormal (power-law-like) distribution: many people earn little, but a non-negligible fraction earns orders of magnitude more.
  • Method described
    • Construct an income distribution for each country.
    • Plot on a logarithmic scale (take logarithms / use log-log style plotting).
    • Obtain an approximately straight line, indicating a power law.
  • Interpretation
    • On a log scale, repeated doublings follow a consistent scaling rule (power-law exponent ~ −1.5 in the narration).

• Lognormal distribution from multiplicative randomness

  • When a quantity evolves by multiplication of random factors (e.g., wealth growing by random % each year), the logarithm converts multiplication into addition.
  • This yields a lognormal payout/wealth distribution.
  • Produces a long tail and inequality, but (as described) it is not the same as a true power-law tail.

• St. Petersburg paradox → power-law payout tail

  • A “doubling until first head” casino game leads to:
    • Infinite theoretical expected value (in the idealized model).
    • A payout distribution with an unbounded tail across many orders of magnitude.
  • When axes are logarithmic, the result becomes a straight line → consistent with a power law.
  • Consequence emphasized
    • No finite variance / effectively infinite standard deviation (in the ideal model).
    • Increasing sample size can shift the estimated mean (extreme events dominate).

• Self-similarity and fractals

  • The narration links power-law behavior to fractal-like, self-similar structure and scale invariance.
  • Examples given: leaf veins, river networks, lung blood vessels, lightning.
  • Mechanistic claim: zooming into the process tree shows repeating structure at smaller scales (fractal analogy).

• Critical phenomena in magnets (phase transitions)

  • Magnetism disappears above the Curie temperature due to loss of alignment of atomic magnetic moments.
  • At the critical point, the system becomes scale-free:
    • Domain-size distributions become power laws.
    • The “internal scale” effectively disappears at criticality.
  • Correlation length / influence range grows:
    • Local interactions near criticality propagate through larger portions of the material.
    • The system becomes highly unstable and less predictable at criticality.

• Self-organized criticality (SOC) in wildfire

  • Wildfire sizes follow power-law-like distributions due to the forest evolving into a critical state via feedback:
    • Sparse areas (few prior burns) → fuel accumulates → large fires possible.
    • Very dense fuel → system becomes primed for large outbreaks.
    • Fires reset local fuel and allow regrowth, returning the system toward criticality.
  • Example narrative:
    • 1988 Yellowstone-area megafires from many ignitions merging into a vast burned region.
  • Simulation described:
    • A grid of cells representing forest/tree presence and growth.
    • Random lightning strikes ignite trees; trees regrow over time.
    • System behavior naturally evolves toward a critical regime.

• Criticality and earthquakes

  • Earthquake magnitude/energy release described as scale-spanning and power-law-like.
  • Example event: Kobe, Japan (1995, Nojima fault) with cascading rupture and large damage.
  • Core claim:
    • Earthquake systems are naturally driven toward critical behavior by stress accumulation and release.

• Sandpile model (avalanche dynamics) as a toy model of criticality

  • Per Bak (and collaborators): a thought experiment/simulation of dropping sand grains onto a grid.
  • Avalanches occur when the pile exceeds local stability rules.
  • Key outcome:
    • Avalanche sizes follow power-law statistics resembling earthquake energy release patterns.
  • Mentioned controversy:
    • Real sand piles do not exactly match the model’s perfect power-law behavior, but the theory is framed as pointing to a universal mechanism.

• Universality

  • Different systems can share the same large-scale statistical behavior near criticality regardless of microscopic details.
  • “Universality classes”:
    • Magnets at specific transitions, fluids, SOC systems (fires/sandpiles/earthquakes), etc.
  • Implication emphasized:
    • Simple models can predict complex systems’ statistical laws.

• Why risk management differs under power laws

  • In power-law environments:
    • Many small events can create false security.
    • Catastrophic extreme events remain likely relative to normal-distribution assumptions.
  • Example applied domain:
    • Insurance and industry underestimation of rare extreme wildfire events (Paradise, CA, 2018 described).
  • Venture capital / media industries:
    • Performance is dominated by a few extreme “breakout” outcomes.

• Network growth and preferential attachment (power laws in the Internet)

  • Albert-László Barabási (early 2000s) studied the web and found no normal distribution of page views; instead a power law emerges.
  • Hypothesis described:
    • New pages are more likely to link to already-popular pages (preferential attachment).
  • Simulation described:
    • Start with a small network.
    • Add nodes over time; each new node connects preferentially to high-degree/popularity nodes.
    • Result: a power-law distribution with exponent roughly −2 (as stated).

• Synthesis / behavioral lesson

  • If a system is normal-distribution-like → averages dominate → consistency matters.
  • If a system is power-law-like → extremes dominate → persistence/risk-taking strategy differs:
    • Make many smart bets hoping one rare outcome dominates.

---

Methodology / simulations mentioned (outline)

• Pareto analysis (income)

  • Collect income data from tax returns (Italy, England, France, Prussia, etc.).
  • Build income distributions per country.
  • Apply logarithmic plotting to reveal a straight-line scaling → infer power-law behavior.

• Casino probability games (illustrative models)

  • Game 1 (additive):
    • 100 fair coin tosses; win $1 per head → expected value computed.
  • Game 2 (multiplicative):
    • Start $1; multiply by factors (1.1 for heads, 0.9 for tails) across 100 tosses → demonstrates lognormal structure.
  • Game 3 (doubling until first head):
    • Payoff doubles until first head; repeated trials yield St. Petersburg paradox behavior → power-law tail.

• Wildfire simulator (self-organized criticality)

  • Use a grid of cells.
  • Each cell can have: no tree / tree / tree growth state.
  • Lightning strikes occur with a probability increasing with an external parameter.
  • Trees grow over time; lightning ignites; fires burn until fuel is consumed.
  • Measure resulting fire sizes; observe power-law-like scaling.

• Sandpile model

  • Initialize a grid and repeatedly drop sand grains (either in the center or randomly).
  • Apply local collapse/instability rules when height exceeds a threshold.
  • Track avalanche sizes and fit frequency distributions → power-law.

• Internet preferential attachment simulation

  • Start with a small network.
  • Add nodes sequentially.
  • Each new node links preferentially to already highly connected nodes.
  • Observe emergent power-law degree/view distribution (exponent ~ −2 in the narration).

---

Researchers / sources featured (as named in the subtitles)

• Vilfredo Pareto
• Abraham De Moivre (subtitles: “Abram Deavr”)
• Per Bak (subtitles: “Perbeck”)
• Kaspar (team member mentioned during magnet simulation; not fully identified in subtitles)
• Albert László Barabási (subtitles: “Albert László Barabászy”)
• Horslee Bridge (investment firm mentioned; spelling as given—likely a source/firm rather than an individual)

Historical/real-world references (named events/policies)

• Yellowstone (1988 wildfire context)
• Kobe, Japan (Nojima fault; January 17, 1995)
• Nojima fault
• U.S. Forest Service and the 10 a.m. policy (1935 policy)
• Paradise, California (2018 wildfire context)
• Bloomsbury (Harry Potter publishing example)
• Netflix / YouTube (viewership distribution examples)
• Delaware and Manhattan (size comparisons)
• Bill Gates, Elon Musk (average-fortune analogy)

(No additional credited publications beyond these are explicitly named in the subtitles.)