SJMcCormick

Statistical thinking will one day be as necessary for efficient citizenship as the ability to read and write.

Samuel S. Wilks

In investing, outcomes are uncertain. You make decisions with partial information and only learn later whether they worked.

That makes probability more useful than prediction. The question is not what will happen, but whether a decision makes sense across many possible outcomes. Over time, the difference between the two compounds.

Process Over Outcome

There’s a common tendency to judge decisions by how they turn out. A good outcome is taken as proof of skill. A bad one is treated as a mistake.

That’s misleading.

A poor decision can still work out because of luck. A sound decision can fail because of variance. Outcomes alone don’t tell you much.

A better way to think about this is expected value (EV) - the probability-weighted average of all possible outcomes. Over repeated decisions, results tend to drift toward EV, but the path matters.

Consider a simple example:

Flip a coin.
Heads, you double your net worth.
Tails, you lose everything.

The expected value is zero, but the downside ends the game. You’ve taken on ruin risk. That violates the first principle of compounding: don’t get knocked out.

Ergodicity and Why the Casino Isn’t Gambling

People often ask whether investing is just another form of gambling. It’s a fair question, but it misses an important distinction.

Individual outcomes are uncertain. Processes are not.

The casino doesn’t win every spin of the roulette wheel. It wins over time. The edge is small, but it’s repeated under conditions where losses never wipe out the house. That’s ergodicity in action.

Time averages diverge from ensemble averages if ruin is possible.

A gambler might get lucky in the short run, but repeated exposure to large losses eventually catches up. The casino sets up a system where outcomes average out in its favour, precisely because survival is never in doubt.

Survival Beats Brilliance

Each realised outcome is only one path among many that could have unfolded. Nassim Taleb calls these “alternative histories”. Howard Marks makes the same point in different language.

Expected value matters. So does how the losses show up (the distribution), and whether you survive them.

You can make a bet that looks attractive on average and still go broke if the left tail is severe enough.

“Don’t be the six-foot man who drowned in the river that was five feet deep on average.”

This is where the Kelly criterion becomes relevant. It formalises the idea of sizing bets to maximise long-term growth while avoiding ruin.

Kelly works under ergodic conditions, where outcomes can be repeated and capital isn’t wiped out along the way. In many real-world settings, especially venture capital and highly concentrated strategies, those assumptions break down. The left tail isn’t just unpleasant. It’s terminal.

Asymmetry Is the Game

This leads to one of the most useful mental models in investing:

“Heads I win, tails I don’t lose much.” – Mohnish Pabrai

The goal is not to win often. It’s to structure decisions where the upside meaningfully outweighs the downside.

This shows up most clearly in venture capital. Most investments fail. A small number of outliers drive the bulk of returns. A single large winner can offset many losses, but only if the fund survives long enough to capture it.

That’s why position sizing and follow-ons matter more than clever entry points.

Reflexivity (Briefly)

In some domains, expectations influence outcomes. This is the idea George Soros describes as reflexivity.

Markets don’t just reflect reality. They help shape it. Investor beliefs affect funding. Funding affects runway. Runway affects outcomes, which then feed back into valuation.

This doesn’t replace probability. It tells you when the probabilities themselves are moving.