Probability and odds are not the same statement
Probability is favourable outcomes divided by total outcomes, expressed from zero to one. Odds compare favourable to unfavourable outcomes. A one-in-four chance is a probability of 0.25 and odds of one to three. The same fact stated two ways, and confusing them is how betting markets are routinely misread.
Converting between them is simple: odds of a to b give a probability of a divided by (a plus b). Probability p gives odds of p to (1 minus p).
Bookmakers quote odds that imply probabilities summing to more than one hundred per cent. The excess is the margin, and it is why the implied probabilities in a market are not the market's honest estimate of what will happen.
Independent events have no memory
A fair coin that has landed heads nine times running still has a fifty per cent chance of heads on the tenth flip. The coin has no mechanism for remembering, and nothing about the previous flips changes the physics of the next one.
The gambler's fallacy, believing a run makes the opposite outcome due, persists because sequences of independent events look non-random to humans. We are pattern-detecting animals and genuinely random data contains more streaks than intuition expects.
The inverse error is just as common: believing a streak indicates a hot hand and will continue. Both errors read structure into a sequence that has none, and both cost money in the same circumstances.
Dependent events and conditional probability
When one event changes the probability of another, the events are dependent. Drawing a card without replacing it changes the deck for the next draw; drawing with replacement does not.
Conditional probability is written as the probability of A given B, and the combined probability of two dependent events is the probability of the first multiplied by the probability of the second given the first. Treating dependent events as independent is the most common error in everyday risk estimation.
It matters in ordinary life more than card games. Component failures in a system are often correlated by a shared cause; insurance claims cluster around the same storm. Multiplying independent probabilities in those situations dramatically understates the chance of everything going wrong at once.
Base rates, and why an accurate test can mislead
Suppose a condition affects one person in a thousand and a test is ninety-nine per cent accurate in both directions. Test ten thousand people: ten have the condition and about ten test positive correctly, while of the 9,990 without it, about a hundred test positive falsely.
So roughly a hundred and ten positive results appear, and only ten are real. A positive result means about a nine per cent chance of having the condition, not ninety-nine. Nothing about the test is faulty; the base rate does all the work.
This is why screening programmes target higher-risk groups, why a positive result on a rare condition is followed by confirmatory testing, and why the same reasoning applies far outside medicine. To fraud detection, security alerts and any rare-event classifier.
Expected value, and where it stops being the whole answer
Expected value is the sum of each outcome multiplied by its probability. A lottery ticket costing two dollars with a one in fourteen million chance of six million has an expected value of about forty-three cents, which is why the ticket is not a financial decision.
Expected value is the right guide when an event repeats many times, because averages converge. It is a poor guide for single, large, irreversible outcomes: a bet with positive expected value that can bankrupt you is still a bad bet, because you do not get to run the experiment again.
This is the practical case for insurance, which has negative expected value by construction, the insurer must profit, and is still rational. You are paying to remove an outcome you cannot absorb, not to win on average.
Questions people ask
- What is the difference between probability and odds?
- Probability compares favourable outcomes to all outcomes; odds compare favourable to unfavourable. A one-in-four chance is a probability of 0.25 and odds of one to three. Converting is straightforward: odds of a to b mean a probability of a divided by a plus b.
- Is the gambler's fallacy really a fallacy?
- Yes, for genuinely independent events. A coin, a roulette wheel and a lottery draw have no memory of previous results. The fallacy feels compelling because random sequences contain longer streaks than intuition expects, so a run looks like evidence of something when it is just what randomness looks like.
- Why can a 99% accurate test still be wrong most of the time?
- Because of the base rate. When a condition is rare, the small percentage of false positives drawn from the large healthy population outnumbers the true positives from the tiny affected group. For a condition affecting one in a thousand, a ninety-nine per cent accurate test produces roughly ten true positives and a hundred false ones.
- How do I calculate the probability of two events both happening?
- If they are independent, multiply their probabilities. If the second depends on the first, multiply the first by the conditional probability of the second given the first. The common mistake is assuming independence when a shared cause links the two, which understates the chance of both occurring.
- Is expected value always the right way to decide?
- Only when the decision repeats enough times for averages to assert themselves. For a single large irreversible outcome it can be badly misleading. Which is why insurance, with its negative expected value, is still rational. You buy it to remove a loss you could not absorb, not to profit on average.
Sources
- Australian Bureau of Statistics: statistical language. Plain-language reference for statistical concepts