Taking Asymmetric Bets

March 24, 2026

The previous piece established a claim: the top of the distribution is underpriced relative to the middle, because fear, information asymmetry, resource constraints, and social norms push people away from high-value, low-probability goals. This piece is about how to capture that.

1. The Mechanism Is Volume

If the value of a high-value, low-probability goal is favorable relative to its cost, the straightforward response is to take the bet repeatedly. Most attempts will fail. This is expected. The math works over volume.

A 5% probability of success means that over 20 attempts, the expected number of successes is 1. Over 100 attempts, it is 5. The individual bet is likely to fail. The portfolio of bets is likely to produce at least one success. The distinction between these two statements is the entire argument.

This is Taleb’s core contribution, and I think he is correct about it. In a world where value is fat-tailed and the cost of each attempt is bounded, repeated asymmetric bets are favorable even when the majority of them fail.

2. The Constraint Is Cost

Volume only works if each bet is cheap enough to take repeatedly. The constraint is not probability. It is the cost of each attempt.

Some bets are cheap. An application takes a few hours. A cold email takes minutes. A side project can be scoped to a few weeks. Publishing something takes an afternoon. For bets at this cost, you can sustain high volume. Low probability is fine because you can take hundreds of these per year.

Some bets are expensive. Training full-time for a national qualification takes years. Relocating for a role has real logistical cost. Building a company consumes everything. For bets at this cost, you cannot sustain volume. You might take one or two in a lifetime. Low probability is much harder to justify, not because the value is necessarily worse, but because you cannot take enough attempts for the probability to resolve.

The practical implication: minimize the cost per bet wherever possible. The cheaper each attempt, the more attempts you can make, the lower the per-bet probability you can tolerate. This is not about cutting corners. It is about structuring your pursuit so that failure on any single attempt does not prevent you from taking the next one.

3. Cheap Bets vs Expensive Bets

This creates a natural categorization.

Cheap bets are attempts where the primary cost is time measured in hours or days, and failure leaves you in approximately the same position as before. Applications, outreach, pitches, small projects, publishing. You can run many of these in parallel. The opportunity cost of any individual failure is negligible.

Expensive bets are attempts where the primary cost is time measured in months or years, and failure leaves you with significant sunk cost. Multi-year training programs, graduate school, building a company, committing to a specific career path. You can run very few of these. The opportunity cost of failure is real.

The interesting cases are bets that look expensive but can be restructured to be cheap. A multi-year training commitment is expensive. But if the training itself develops skills that transfer regardless of whether you hit the specific threshold (health, discipline, credibility), the sunk cost on failure is lower than it appears. The bet is partially hedged by its side effects.

Conversely, some bets look cheap but are actually expensive. The gym is a good example. One session costs an hour. That feels cheap. But the value curve is smooth and flattening. There is no threshold where it jumps. Hundreds of hours for marginal, diminishing returns. The cumulative cost relative to the value produced is high. It just doesn’t feel that way because the per-session cost is low.

This is why “cost” as a concept needs to account for cumulative investment relative to the shape of the value curve. A bet is expensive if the total time spent is high relative to the value it produces, regardless of how cheap each individual session feels.

4. The Portfolio

If you accept the claim from the previous piece (the top is underpriced, the middle is overpriced), the portfolio construction follows:

Maximize the number of cheap, high-value bets. These are attempts at the top of the distribution that cost little per attempt. Applications to the best positions, not the attainable ones. Outreach to the most interesting people, not the most accessible ones. Projects aimed at the highest-impact outcome, not the most likely one.

Be deliberate about expensive bets. These cannot be taken in volume, so the calculus is different. An expensive bet is worth taking when the value at the threshold is so large that even one or two attempts at that probability justify the cost, or when failure has significant hedging (the process itself produces transferable value).

Avoid the middle. This is the most counterintuitive part. The attainable-feeling goals with 25-30% probability and mid-high value are where competition is most disproportionate. Effort spent here has worse expected return per unit than effort spent at the top (cheap bets, high value) or the bottom (low effort, honest value, no competition).

5. What This Does Not Address

This framework is mechanical. It treats goal selection as a portfolio problem. Given the structure of value, given the cost constraints, here is how to allocate.

What it does not address is why people fail to execute this even when they understand it. The answer, as I noted in the previous piece, is primarily psychological. Fear of failure is not an information problem. Knowing the math does not make you less afraid.

That is a different problem, and it is treated separately.