Contents
Mathematics & Probability
Section 1
The Core Idea
Randomness is the absence of a deterministic pattern: outcomes that cannot be predicted with certainty from prior information. In practice, we treat a process as random when we lack the information or the model to predict it — or when the process is inherently stochastic (e.g. quantum, thermal noise). The strategic implication is that in random or near-random domains, you cannot infer cause from a single outcome, you cannot expect "reversion" on the next trial, and you should not overfit narratives to short runs. Deciding and judging under randomness require probabilistic thinking: think in distributions and base rates, not in single outcomes and stories.
The human bias is to see pattern where there is none. We attribute a streak of wins to skill and a streak of losses to bad luck; we treat the last few data points as predictive of the next. In truly random processes, past outcomes do not change the probability of future ones (see Monte Carlo fallacy). The discipline is to ask: is this process random or does it have structure? If random (or effectively so), use base rates and distributions; don't overinterpret short sequences. If there is structure (skill, causality), then history can inform the future — but the burden of proof is on showing the structure.
Randomness also separates signal from noise. When outcomes are partly random, the noise can obscure the signal (e.g. true skill, true quality). Understanding and analysing require distinguishing what is repeatable from what is chance. Use the model to calibrate confidence: in random domains, be humble about prediction; in structured domains, invest in learning the structure.
Section 2
How to See It
Randomness shows up when outcomes vary in ways that don't seem fully explained by known factors, when people disagree whether a result was "luck" or "skill," or when short runs are overinterpreted. The diagnostic: could this outcome have occurred by chance given the base rate? Is there evidence of structure (repeatable skill, causal mechanism) or are we seeing noise?
Business
You're seeing Randomness when a sales rep has a great quarter and is credited with "doing something right," then has a bad quarter. If deal outcomes are partly random (timing, buyer mood, competition), the run may be noise. Deciding and judging require asking whether performance is repeatable or within the range of chance.
Investing
You're seeing Randomness when a fund beats the market three years in a row. The question is whether that is skill (structure) or luck (randomness). Without enough trials or a clear edge, we cannot tell. Understanding & analysing require distinguishing alpha from random variation.
Product
You're seeing Randomness when A/B test results flip with a few more days of data. The outcome is noisy; the "winner" was within the margin of randomness. Deciding requires enough sample size and a probabilistic view (confidence intervals, not point estimates) so that randomness doesn't drive false conclusions.
Strategy
You're seeing Randomness when a strategy works in one market and fails in another. Was the first success causal (we did X and got Y) or partly random (we happened to be in the right place)? Understanding requires asking what would have happened under different random draws — and whether the strategy is robust to noise.
Section 3
How to Use It
Decision filter
"When evaluating an outcome or a run of outcomes, ask: how much of this could be randomness? Use base rates and distributions; don't attribute cause or skill without evidence of structure. In deciding, think in probabilities; in analysing, separate signal from noise. Calibrate confidence to the role of chance."
As a founder
Don't overinterpret short runs — one great month, one bad hire, one lost deal. Ask whether the outcome is within the range of chance given your base rates. Use probabilistic thinking for forecasts and for evaluating people and strategies: require more data or clearer causal evidence before concluding "this works" or "this doesn't." When presenting results, acknowledge randomness (e.g. confidence intervals, ranges) so decisions aren't driven by noise.
As an investor
Size and evaluate strategies with randomness in mind. Past returns may be luck. Ask for process, edge, and sample size — not just outcome. When outcomes are partly random, diversify and size positions so that one random draw doesn't determine fate. Separate alpha from noise before paying for performance.
As a decision-maker
When someone presents a single outcome or a short run as evidence, ask: what is the base rate? What would we expect by chance? Require probabilistic framing (e.g. "70% confidence," "consistent with random") when the process has material randomness. Resist narrative explanations that ignore the role of chance.
Common misapplication: Treating all variation as randomness. Some outcomes are causal — we did X and got Y. The discipline is to test for structure (e.g. repeatability, mechanism) and to use randomness as the default when we cannot show structure. Don't assume randomness when there is evidence of cause; don't assume cause when there is only a short run.
Second misapplication: Using "it's random" to avoid analysis. Randomness doesn't mean "we can't learn." It means single outcomes are weak evidence. We can still estimate base rates, run experiments, and build models that account for noise. The point is to not overinterpret the single draw.
Section 4
The Mechanism
Section 5
Founders & Leaders in Action
Ed ThorpMathematician, Beat the Dealer; hedge fund founder
Thorp's work on blackjack and later on markets explicitly separated random processes (e.g. roulette, fair dice) from those with structure (e.g. card counting). He emphasised knowing which is which and thinking in probabilities — the core of deciding and judging under randomness.
Charlie MungerVice Chairman, Berkshire Hathaway
Munger repeatedly warns against attributing outcomes to skill or cause when luck could explain them. He stresses base rates, "invert always invert," and not being fooled by short runs. His mental models are aimed at separating signal from noise and avoiding narrative fallacy in the face of randomness.
Section 6
Visual Explanation
Randomness — Outcomes as draws from a distribution. Don't overinterpret single outcomes; use base rates and distributions. Separate signal from noise.
Section 7
Connected Models
Randomness sits with probability, judgment, and signal. The models below reinforce it, create tension, or extend into practice.
Reinforces
Probability Theory
Probability theory is the mathematical framework for randomness: distributions, expectation, independence. Randomness is the domain where probability applies. The two are inseparable — when we say "random," we mean "described by a probability distribution."
Reinforces
Law of Large Numbers
The law of large numbers says that sample averages converge to the expected value as the number of trials grows. So under randomness, we learn from many trials, not from one. The LLN justifies using base rates and long-run statistics; it does not justify expecting the next outcome to "correct" for the past (that's the Monte Carlo fallacy).
Tension
Monte Carlo Fallacy
The Monte Carlo fallacy is expecting the next random outcome to "revert" or "balance" the past (e.g. "we're due"). Under true randomness, the next outcome is independent. The tension: we want to learn from history (LLN over many trials) but not misapply that to the next single draw.
Tension
[Narrative](/mental-models/narrative) Fallacy
The narrative fallacy is constructing a story that makes sense of outcomes after the fact. When outcomes are partly random, the story can be overfit to noise. The tension: we crave narrative; randomness says single outcomes are weak evidence. Resist narrative when the process is random.
Leads-to
Probabilistic Thinking
Probabilistic thinking is reasoning in terms of probabilities, base rates, and distributions rather than single outcomes and certainty. Randomness is the context where that discipline is essential. When we accept that outcomes are random (or partly so), we are led to probabilistic thinking for deciding and judging.
Leads-to
Signal vs Noise
When outcomes are partly random, the "noise" is the random component; the "signal" is the repeatable component (skill, cause). Understanding & analysing require separating the two. Randomness leads directly to the question: how much of this outcome is signal vs noise?
Section 8
One Key Quote
— Pierre-Simon Laplace, A Philosophical Essay on Probabilities"Chance is nothing but the expression of our ignorance of the causes of events."
Laplace's formulation is epistemic: we call something "random" when we don't have the information or the model to predict it. That doesn't mean nothing causes the outcome; it means we treat it as a draw from a distribution. The discipline is to act on that distribution (base rates, probabilities) rather than pretending we know the cause of a single outcome.
Section 9
Analyst's Take
Faster Than Normal — Editorial View
Most misjudgments under randomness come from overinterpreting single outcomes. We see a win and attribute it to skill; we see a loss and attribute it to bad luck. In random or partly random domains, single outcomes are weak evidence. Require base rates, sample size, or causal mechanism before concluding.
Deciding and judging both need probabilistic calibration. When the process has material randomness, state confidence in ranges and probabilities — not in certainties. "We have a 70% chance of hitting the target" is better than "we will hit the target" when outcomes are noisy. Calibrate your language to the role of chance.
Separate signal from noise before acting. When outcomes are partly random, the repeatable part (signal) is what you can improve or bet on; the noise is what you must diversify or accept. Understanding & analysing require that separation. Don't build strategy on noise.
Section 10
Test Yourself
Is this mental model at work here?
Scenario 1
A salesperson has three great quarters in a row. Management concludes they have 'cracked the code' and rolls out their method to the team.
Scenario 2
A coin is flipped 10 times; 7 heads. Someone says 'this coin is biased toward heads.'
Scenario 3
An investor says 'I only invest when I have high conviction — I need to understand the cause of past returns.'
Scenario 4
A team runs an A/B test for one week; variant B wins 52% to 48%. They ship B.
Section 11
Summary & Further Reading
Summary: Randomness is the absence of deterministic predictability; we describe such outcomes with distributions and probabilities. In deciding and judging, don't overinterpret single outcomes or short runs; use base rates and probabilistic thinking. In understanding and analysing, separate signal (repeatable) from noise (random). Calibrate confidence to the role of chance. Pair with probability theory, law of large numbers, Monte Carlo fallacy, and narrative fallacy; extend to probabilistic thinking and signal vs noise.
Further Reading
01
Book
Kahneman on how we misread randomness: we see pattern, attribute cause, and underweight base rates. Central to deciding and judging under uncertainty.
02
Book
How randomness shapes outcomes in life and markets. Accessible treatment of when we mistake luck for skill and vice versa.
03
Book
Taleb on the role of chance in success and failure. Why we attribute too much to skill and too little to randomness.
04
Book
Laplace's framing of probability and chance. Foundational for the epistemic view of randomness.
05
Book
On probabilistic forecasting and calibration. How to think in probabilities when outcomes are uncertain and partly random.
Why this matters next
mental modelsNarrative
Randomness applied the Narrative mental model
mental modelsQuality
Randomness applied the Quality mental model
mental modelsMonte Carlo Fallacy
Randomness applied the Monte Carlo Fallacy mental model
mental modelsVariance
Randomness applied the Variance mental model
mental modelsUncertainty
Randomness applied the Uncertainty mental model
mental modelsDistribution
Randomness applied the Distribution mental model
Frequently asked questions
What is Randomness?
Randomness is a mental model used for better thinking and decision-making.
How do you apply Randomness?
To apply Randomness, identify situations where this framework is relevant, then use it as a lens to evaluate your options and decisions. The model is most useful when combined with other complementary mental models.
What category does Randomness fall under?
Randomness falls under the Mathematics & Probability category of mental models. Other models in this category can be found on the Mathematics & Probability hub page.
Why is Randomness important?
Randomness is important because it provides a structured way to think about problems that would otherwise be approached with intuition alone. Understanding this model helps you avoid common reasoning errors and make better decisions.
Continue exploring
BTBTCOCCEREG
Mental model
Bayes Theorem
A mathematical framework for updating beliefs based on new evidence, proportiona
Mental model
Black Swan Theory
Rare, unpredictable events with extreme impact that are retrospectively rational
Mental model
Compounding
Small consistent gains accumulate exponentially over time — the most powerful fo
Mental model
Correlation vs Causation
The critical distinction between two variables that move together and one actual
Mental model
Ergodicity
The distinction between ensemble averages and time averages — what works across
Mental model
Exponential Growth
Growth that accelerates proportionally to its current size, producing deceptivel
More like this, in your inbox
I send a newsletter every week — free, no spam, unsubscribe anytime.