Probability Guide 2026: Rules, Odds and Real Examples
Table of Contents
Basic Rules
Coin flip heads twice: 0.5 x 0.5 = 0.25.
Die shows 1 or 6: 1/6 + 1/6 = 1/3.
NOT: P(not A) = 1 - P(A). At least one head in 3 flips: 1 - (0.5)^3 = 0.875.
Drawing without replacement changes the denominators as you go. The chance of two aces from a shuffled deck: (4/52) x (3/51) = 12/2652, about 0.45%. The second draw has only 51 cards and 3 aces left - the events are dependent, so each fraction reflects the shrunken deck. Card counters live in these shifting denominators.
The complement trick solved probability's founding bet. De Mere asked: at least one six in four die rolls? Compute the opposite - no sixes: (5/6)^4 = 0.4823 - and subtract from 1: 0.5177. Slightly better than 50/50, which is why the bet was profitable and why Pascal and Fermat's 1654 correspondence launched probability theory. "At least one" almost always means "one minus none."
Conditional Probability
Cards: P(ace | red card drawn)? P(ace and red) = 2/52, P(red) = 26/52. Result: (2/52)/(26/52) = 2/26 = 1/13.
Medical testing: A 99%-accurate test for a 1-in-1000 disease: positive result means only ~9% chance of disease (base rate dominates). This trips up even doctors.
The Monty Hall problem is conditional probability's most famous ambush. You pick door 1 of 3; the host, who knows where the car is, opens a goat door; switching wins 2/3 of the time. Your first pick was right with probability 1/3, and that never changes - the host's forced move concentrates the remaining 2/3 onto the unopened door. Condition on what the host's knowledge forced him to do, not on your gut.
Hiring tests run into the same base-rate wall as medical tests. Suppose 10% of applicants are truly qualified; the test passes 90% of qualified candidates but also 20% of unqualified ones. P(qualified | pass) = (0.9 x 0.1)/(0.9 x 0.1 + 0.2 x 0.9) = 0.09/0.27 = 33.3%. Two-thirds of "passing" candidates are unqualified - the low base rate dominates the decent test. Screen with base rates in view or drown in false positives.
Odds vs Probability
| Format | Example | Conversion |
|---|---|---|
| Probability | 0.25 | Odds = p/(1-p) = 1:3 |
| Odds (against) | 3 to 1 | p = 1/(3+1) = 0.25 |
| American (+200) | Bet $100 wins $200 | p = 100/(200+100) = 33.3% |
Fractional odds need conversion before comparison. Odds of 5/2 ("five to two") mean probability 2/(5+2) = 28.6%, and a winning $100 bet returns $250 profit. UK horse racing speaks fractions, US sportsbooks speak moneylines (+250 is the same bet), and exchanges speak decimals (3.50). Convert everything to implied probability first - only then can you compare prices across bookmakers.
Implied probabilities that sum past 100% reveal the house edge. Both sides of a game priced at -110 carry implied probability 52.4% each, totaling 104.8%. That extra 4.8% is the vigorish - the bookmaker's margin, guaranteeing profit regardless of outcome. Any market whose probabilities sum above 100% is charging you for the privilege of betting.
Famous Probabilities
- Powerball jackpot: 1 in 292,201,338. Buying 100 tickets barely moves it.
- Birthday paradox: 23 people: 50% chance of shared birthday. 70 people: 99.9%.
- Perfect bracket: 1 in 9.2 quintillion (coin flip); ~1 in 120 billion with basketball knowledge.
- Coin flip streaks: 10 heads in a row: 1 in 1,024 - happens regularly across millions of flippers.
The lightning benchmark puts rare events in perspective. Your yearly chance of being struck by lightning in the US is about 1 in 1.22 million - roughly 240 times likelier than winning the Powerball jackpot. Risk communicators use lightning as the yardstick: if a hazard is "less likely than a lightning strike," it is negligible for practical purposes. Keep the comparison handy for lottery tickets and shark attacks alike.
The gambler's fallacy dies hard because streaks feel meaningful. After eight reds on a roulette wheel, the next spin is still 18/38 - about 47.4% - red. The wheel has no memory, and "due for black" is a story the brain tells, not a probability. Martingale bettors do not lose to bad odds; they lose to table limits that cap the doubling sequence.
Frequently Asked Questions
Probability = favorable/total outcomes (0 to 1). Odds = favorable:unfavorable. 25% probability = 1:3 odds.
Independent AND: multiply. Mutually exclusive OR: add. 'At least one': 1 minus P(none).
P(A|B): probability of A given B happened. Formula: P(A and B)/P(B). Essential for medical tests and Bayesian reasoning.
We think in terms of matching OUR birthday (1/365 each), but the paradox counts ANY pair matching - with 23 people there are 253 pairs.
Probability-weighted average outcome: sum(p x payoff). Positive EV bets profit long-term; lotteries have deeply negative EV.