Chapter 3 Exercises: Randomness Is Real

These exercises build from conceptual comprehension (Level 1) through analysis, application, synthesis, and original research (Level 5). Work through them in order — each level builds on the last.

Level 1: Comprehension — Getting the Basics

These questions check that you understood the key ideas. If you can't answer them from memory, review the relevant chapter sections before moving on.

Exercise 1.1 — The Two Types of Unpredictability

In your own words (two to three sentences each), explain the difference between: - Practical unpredictability (deterministic chaos) - True randomness (ontological/intrinsic randomness)

Then give one real-world example of each type. Do not use the examples given in the chapter.

Exercise 1.2 — The Coin Flip Illusion

A friend flips a coin five times and gets: H, H, H, H, H.

Your friend says: "This coin must be biased. Five heads in a row can't happen with a fair coin."

a. Is your friend right? Calculate the probability of getting five heads in a row with a fair coin. Show your reasoning (you don't need to know formal probability formulas — reason it through step by step).

b. What cognitive error, introduced in the chapter, does your friend's statement illustrate?

c. How many flips would you need to observe before you could be reasonably confident the coin is biased (not perfectly confident — just reasonably)? Explain your thinking.

Exercise 1.3 — Individual vs. Ensemble Prediction

For each of the following scenarios, classify the type of prediction being made (individual or ensemble) and explain whether randomness makes it more or less reliable:

a. "This specific startup, launched this year, will succeed." b. "Roughly 90% of new restaurants close within five years." c. "Nadia's next video will get more than 10,000 views." d. "A content creator who posts daily for two years has a higher expected viewership than one who posts twice a month." e. "It will rain somewhere in Seattle during October."

Exercise 1.4 — The Salganik/Watts Music Lab

Answer the following questions about the music lab study described in the Research Spotlight:

a. What was the key experimental manipulation — what did the "social influence" condition change compared to the "independent" condition?

b. Why did the researchers run eight parallel worlds rather than just two (independent vs. social influence)?

c. What does the finding mean for our intuitive explanations of why some songs become hits? Be specific.

d. Name two domains outside of music where you would expect this same dynamic to operate.

Exercise 1.5 — Vocabulary Check

Define each of the following terms from the chapter, and give one example of each that was not used in the chapter:

• Apophenia
• The gambler's fallacy
• Cumulative advantage
• Regression to the mean (preview definition — use what was given in the chapter)
• Ensemble prediction
• Path dependence

Level 2: Analysis — Looking Under the Hood

These exercises ask you to examine arguments, identify assumptions, and evaluate evidence.

Exercise 2.1 — Dissecting a Creator's Logic

A YouTuber posts the following analysis of their channel's performance:

"I've figured out the algorithm. In the three months I started using longer video descriptions, my views went up 40%. I've tested this twice — both months I used long descriptions, my views were higher. I'm now advising all my creator friends to use longer descriptions."

Identify at least four specific analytical problems with this creator's reasoning. For each problem, name the concept from the chapter it relates to, and explain what the creator should have done instead.

Exercise 2.2 — Evaluating the Determinism Argument

The hard determinist says: "Every outcome is determined by prior causes following fixed physical laws. Therefore, 'luck' is just a name for our ignorance — it doesn't really exist."

The chapter presents three counter-arguments to this position. Restate each counter-argument in your own words, then evaluate: which of the three do you find most persuasive, and why? Is there a weakness in the most persuasive counter-argument that the chapter doesn't fully acknowledge?

Exercise 2.3 — The Hot Hand Debate

The chapter mentions that newer research suggests there may be a "small genuine hot hand effect" in some sports, complicating the original Gilovich/Vallone/Tversky finding.

a. Why does the existence of a real (if small) hot hand effect not undermine the chapter's main point about pattern-perception in random sequences?

b. In sports like basketball, multiple variables can affect shooting success beyond pure chance (fatigue, defensive pressure, location on court, game situation). How might these variables complicate any attempt to measure a "hot hand" effect? What would a methodologically sound study need to control for?

c. If a small genuine hot hand effect exists (say, players who made their last 3 shots are 2% more likely to make the next one), what practical implications does this have for strategy? Does a 2% effect matter?

Exercise 2.4 — Marcus's Objection, Extended

Marcus argues that some creators are consistently successful over long periods, which he takes as evidence that skill, not luck, determines outcomes. Dr. Yuki agrees — over long periods, skill becomes the dominant signal.

But here is a complication: survivorship bias.

a. How does survivorship bias complicate Marcus's argument? (If you need to, you can read ahead briefly to the survivorship bias summary in Chapter 9's preview materials, but try to reason it from the chapter description first.)

b. If we only observe creators who are still active and successful after five years, what does that tell us about the full population of creators who tried? What can't we conclude from observing this survivors-only population?

c. Construct a scenario where a creator who is "consistently successful over five years" may be predominantly lucky rather than predominantly skilled. What would that look like?

Exercise 2.5 — Mapping Randomness Levels

Different domains have different noise-to-signal ratios — some activities are more luck-dominated than others. Based on what you've learned in this chapter and your general knowledge, order the following activities from most skill-dominated to most luck-dominated and explain your ranking briefly for each:

• A single 5-minute chess game between equally-rated players
• Predicting next month's stock market return
• A 100-game professional chess match between equally-rated players
• Getting a specific song to go viral on TikTok
• Predicting whether a creator with 100K subscribers will reach 1M in 3 years
• The outcome of a single medical clinical trial with n=30

Level 3: Application — Using the Ideas

These exercises ask you to apply chapter concepts to new situations, including your own life.

Exercise 3.1 — Your Own Coin Flip Analysis

Choose one domain in your life where you evaluate your own performance: academic grades, athletic performance, creative work, social media, gaming, or any other domain where you get repeated measurable outcomes.

a. Describe the domain and how you typically measure your performance.

b. Estimate the likely noise-to-signal ratio in this domain based on the principles from this chapter. Is individual outcome variation high or low? What factors introduce randomness?

c. Identify one time when you drew a lesson from a very small sample of results (two or three outcomes). Looking back, was the lesson valid given what you now know about sample size?

d. What would "ensemble thinking" look like applied to your performance in this domain? What would you need to track, over how long, to get a reliable signal?

Exercise 3.2 — The Algorithmic Randomness Audit

If you use any social media platform as a creator, consumer, or both, conduct the following analysis. (If you don't personally create, use a publicly visible creator's account.)

a. Examine at least 15 recent posts/videos/pieces of content and record their engagement metrics (views, likes, shares, or whatever is available).

b. Calculate the mean, highest value, and lowest value. What is the ratio of highest to lowest? This ratio gives you a rough sense of variance.

c. Look at the highest-performing piece. Can you identify a clear causal reason why it performed better? Does that reason also explain the second and third highest performing pieces? If not, what does that suggest?

d. Based on your analysis, estimate what fraction of the outcome variation appears explainable by identifiable factors (quality, topic, timing) vs. what appears to be random noise. Describe your reasoning.

Exercise 3.3 — Ensemble Design Exercise

Priya is job searching. Using the ensemble-thinking framework from this chapter, help her redesign her approach.

She currently does the following: - Spends 3 hours on each application, customizing extensively - Applies to approximately 5 positions per month - Carefully reviews each rejection for "lessons" - Has applied to 47 positions over 9 months

Reframe her approach using what you now know about: - Individual vs. ensemble prediction - Sample size and signal quality - Improving the distribution vs. optimizing for specific outcomes - The role of random "seeding" conditions in selection processes

What specific changes would you recommend, and why?

Exercise 3.4 — Randomness and Moral Attribution

The chapter notes that we attribute our own successes to effort and our failures to bad luck (self-serving attribution bias), while attributing others' successes to luck and their failures to character flaws.

a. Give two specific examples from your own experience where you may have used this asymmetric attribution pattern. Be honest.

b. Now apply the randomness framework from this chapter: how much of each outcome you described was realistically skill/character, and how much was likely random variation?

c. The moral dimension: if outcomes are partly random, how should that affect how we judge ourselves and others for success and failure? Draft a 3–4 sentence personal philosophy on this question.

Exercise 3.5 — The Warhol Problem

Andy Warhol famously predicted that "in the future, everyone will be world-famous for 15 minutes." Interpret this claim using the randomness framework:

a. Does the Salganik/Watts music lab research support the Warhol prediction in any sense? How?

b. What does "15 minutes of fame" look like in the social media era? Has the algorithmic spread of content made Warhol's prediction more or less accurate?

c. If cultural fame is substantially random in the way the music lab experiments suggest, what are the ethical implications for how we think about celebrity culture, the celebrity economy, and how we allocate our attention to famous people?

Level 4: Synthesis — Building New Ideas

These exercises ask you to connect ideas, build arguments, and create original analyses.

Exercise 4.1 — The Signal vs. the Story

In the chapter, we argue that post-hoc explanations of why content went viral are usually wrong — not because the explainer is dumb, but because the event is substantially underdetermined by observable factors.

Write a structured argument (400–500 words) that addresses the following question: If post-hoc explanations are often unreliable, does this mean we should stop trying to analyze what works? Your argument should: - Acknowledge the strongest version of the "stop analyzing" position - Explain why you agree or disagree - Distinguish between analyses that are likely to be useful and those that are likely to generate noise-based lessons - Draw on at least two specific concepts from this chapter

Exercise 4.2 — Designing a Better Performance Review

Many organizations evaluate employee performance using annual reviews that look at outcomes over a one-year window. Using everything you know about randomness, individual vs. ensemble prediction, and signal-to-noise ratios, write a critique of this practice and design a better alternative.

Your alternative should: - Specify what would be measured and over what time horizon - Explain how it would account for the random components of individual performance - Address the challenge of rewarding genuine skill even when lucky and unlucky outcomes are mixed in - Be specific enough to actually implement

Exercise 4.3 — Connecting to Chapter 2

In Chapter 2, we discussed the luck-skill continuum — the idea that different activities fall at different points on a spectrum from pure skill to pure luck.

Now, using the randomness concepts from Chapter 3, develop a more precise account of what "luck-dominated" actually means. Specifically:

a. Propose a working definition of "luck-dominated domain" that uses the language of noise-to-signal ratio, sample size, and ensemble prediction.

b. Using your definition, classify three domains: professional chess, content creation, and venture capital investing. Are your classifications consistent with intuition? Where do they surprise you?

c. What would it mean to "practice" in a luck-dominated domain? Is deliberate practice as valuable in a luck-dominated domain as in a skill-dominated one? Why or why not?

Exercise 4.4 — Writing the Counter-Argument

The chapter argues that randomness is pervasive and that our attribution patterns are systematically wrong. Write the strongest possible counter-argument to this position — a 300–400 word argument defending the view that outcomes in creative and professional domains are much more skill-determined than the chapter acknowledges.

Then respond to your own counter-argument in 200–300 words. Where is it strong? Where does it fail?

Level 5: Research and Extension

These exercises involve going beyond the chapter to engage with primary sources, empirical data, and original creative work.

Exercise 5.1 — Find and Evaluate a Primary Source

The Salganik, Dodds, and Watts (2006) study on music markets ("Experimental Study of Inequality and Unpredictability in an Artificial Cultural Market," Science 311:854–856) is referenced in this chapter.

a. Find and read (or carefully summarize, if you can only access the abstract) the original study. What was the sample size? What were the exact conditions? Were there any limitations the authors acknowledged?

b. One critique of this study is that the "songs" were by unknown artists, which may not be representative of how real celebrity music markets work (where prior fame already creates massive cumulative advantage). Evaluate this critique. Does it substantially change the interpretation?

c. Find one follow-up study or piece of research that either replicates, extends, or challenges the Salganik/Watts findings. Summarize its main argument and your assessment of it.

Exercise 5.2 — The Empirical Hot Hand

The hot hand debate has generated dozens of studies. Your task:

a. Find three published studies on the hot hand phenomenon, from three different sports or domains (e.g., basketball, soccer, horse racing, financial markets, or elsewhere).

b. Summarize each study's method and finding.

c. Synthesize: does the evidence across these domains suggest there is a genuine hot hand effect? Does the effect size matter? Under what conditions does the effect appear vs. disappear?

d. What does this research say about the broader claim in this chapter that human pattern-detection over-perceives persistence in random sequences?

Exercise 5.3 — The Algorithm Experiment

Design (and, optionally, run) a simple personal experiment to test the role of randomness in social media content performance.

Your experiment should: - Specify a platform and what you will post - Hold constant as many variables as possible (quality, length, topic, posting time) while varying one element - Run for at least 20 trials (posts) before drawing conclusions - Pre-specify what result would count as evidence that the manipulated variable does matter vs. evidence that outcomes are predominantly noise - Include how you will control for the possibility that your growing or shrinking audience during the experiment period is itself affecting results

Write up the experimental design in 400–500 words. If you run it, write a brief report of the results.

Exercise 5.4 — Randomness in a Domain You Know

Choose a domain you know personally — a sport, a creative field, an academic subject, a game. Write a 600–800 word analysis of how randomness operates in that domain, drawing on:

• The distinction between individual and ensemble outcomes
• The noise-to-signal ratio in typical performance evaluations in that domain
• How practitioners in that domain typically misattribute random variation
• What a "randomness-aware" approach to skill development in that domain would look like

This should go beyond general statements and include specific examples from your knowledge of the domain.

Answers to selected exercises are in Appendix B.