Round-Trip Fallacy: When P(A|B) Isn't P(B|A)

The round-trip fallacy is not primarily a problem for mathematicians. It's a problem for policy, for life decisions, and for how we reason about causality.

It is the confusion between "Given that X, what's the probability of Y?" and "Given that Y, what's the probability of X?" These are asking about completely different things. They have completely different answers. But your mind treats them as if they were the same question.

The "House Always Wins" Trap.

Here is a version most people carry without noticing. "The house always wins" is folk wisdom, and in a narrow sense it is true: roughly 98% of sports bettors lose over time. From that, almost everyone concludes: "Therefore you cannot profit from sportsbooks." Watch the inversion.

The true statement is about the population: P(loses money | random bettor) is high, around 98%. The conclusion people draw is directional and different: P(you cannot profit | you use a hedging method) — which does not follow at all. The 98% are recreational bettors picking sides on feel. Their loss rate says nothing about a bettor who never predicts an outcome, who instead places two paired bets at two different books so exactly one wins and the profit is fixed before kickoff. A $500 bet on the Lakers at +120 covered by $524 on the Celtics at +110 collects $1,100 either way — $76 of locked profit on a $1,024 outlay, 7.4% in one evening. Sportsbooks themselves disprove the round-trip: if profit were impossible, they would not employ teams to identify winning customers and limit them. You do not build a detection system for something that cannot happen. "Most bettors lose" is true; "you cannot profit" is the fallacy that keeps competition thin.

The Smoking Example

Most smokers do not get lung cancer. This is statistically true.

Someone uses this fact to argue: "Therefore, smoking is not that dangerous."

But wait. Let's be precise about what each claim actually says.

"Most smokers don't get lung cancer" means: P(no cancer | smoker) is high, maybe around 90%. Of all the people who smoke, roughly 90% don't develop lung cancer.

"Smoking isn't dangerous" would mean: the probability of cancer is approximately the same for smokers and non-smokers. P(cancer | smoker) ≈ P(cancer | non-smoker).

These are not the same thing. The first can be true while the second is false — and it is.

Among non-smokers, maybe 5 out of 100 develop lung cancer in a lifetime. Among smokers, maybe 20 out of 100 do. So:

The statement "most smokers don't get cancer" is correct. But it has been used to argue that smoking isn't a major cause of cancer, which is completely wrong.

Smoking overwhelmingly causes lung cancer (P(cancer | smoker) is four times higher than P(cancer | non-smoker)). But most smokers individually don't develop lung cancer because lung cancer is still rare at the population level.

The two conditional probabilities run in opposite directions. Confusing their direction is the round-trip fallacy.

The "Most Terrorists Are Muslim" Trap

This example has enormous real-world policy consequences.

Statement: "Most terrorists are Muslim."

This could be empirically true depending on how you define "terrorist" and what time period you examine. In certain contexts and definitions, many terrorist attacks are carried out by individuals of Muslim faith.

Statement derived by round-trip: "Most Muslims are terrorists."

This is wildly false. Billions of Muslims live peacefully. The probability that a randomly selected Muslim person is a terrorist is vanishingly small — perhaps less than 0.001%.

But because the first statement is true, some people conclude the second must be true, or at least that Muslims-as-a-group should be treated as a security threat.

This is the round-trip fallacy producing policy.

The actual math:

Let's say 1% of terrorist attacks are carried out by Muslims (just for illustration). And let's say there are 2 billion Muslims in the world. And let's say there are 10,000 terrorist attacks per year globally.

A random Muslim person is actually less likely than a random person generally to carry out a terrorist attack, because Muslims are such an enormous population.

But the statement "most terrorists are Muslims" (which could be true) has been inverted into "most Muslims are terrorists" (wildly false) with massive policy consequences.

The Medical Test Paradox

I cover this in more detail elsewhere, but it deserves mentioning here because it is the clearest example of why this matters.

A disease affects 1 in 10,000 people. A test is 99% accurate both ways (99% sensitivity and 99% specificity).

You test positive. What's the probability you have the disease?

The round-trip confusion suggests the answer is 99%. "The test is 99% accurate, so if I test positive, I'm 99% likely to have it."

The actual answer is about 1%.

Here's the math:

Of 10,000 people: - 1 has the disease. The test correctly identifies them (99% sensitivity). So 1 positive. - 9,999 don't have the disease. The test correctly identifies 99% of them as healthy. But 1% are false positives — that's 99 people.

Total positives: 1 (true) + 99 (false) = 100.

Among everyone who tests positive, only 1 is actually sick. You are 99 times more likely to be a false positive.

The round-trip confusion: people confuse P(positive | disease) with P(disease | positive).

This has led to: - Unnecessary medical procedures and anxiety - Lawsuits based on misunderstood DNA test results - Policy decisions based on misunderstanding prevalence of problems

Once you see it, the pattern is everywhere.

How to Protect Yourself

When someone states a statistical fact about a group or a cause-and-effect relationship, ask:

Is this the claim, or the inverse?

"Most successful people work hard" is not the same as "most hard workers are successful."

"People with graduate degrees earn more" is not the same as "earning more causes you to have a graduate degree."

"Convicted criminals are disproportionately likely to have been abused as children" is not the same as "abused children are disproportionately likely to become criminals." (The base rate matters.)

Always ask: which direction is the claim going?

When the direction is inverted in the conversation, you are seeing the round-trip fallacy in action.

The practical move: Always ask for base rates. If someone says a group has a certain property, ask: what fraction of the larger population has that property?

"Most terrorists are Muslim" combined with "Muslims are 25% of the world population" immediately suggests the round-trip is not going the direction people think.

The Ungambled method is built for people who have escaped one specific round-trip fallacy: mistaking "most bettors lose" for "profit is impossible." The direction that actually matters is P(guaranteed profit | you hedge a priced disagreement), and that probability is 100% by construction, because the two locked payouts cover both outcomes. The stigma does the method a favor. Every person who confuses the two conditionals is one fewer competitor for the finite pool of bonuses and price gaps.