A coin and a die are the two most useful teaching tools in probability. Their outcomes are simple, their math is exact, and their behavior contradicts intuition in revealing ways. Streaks of heads, dice that "feel hot," and the gambler's fallacy all live in this small space, and once you understand how random events actually behave, much of probability stops being mysterious.
Key Takeaways
- A fair coin has P(heads) = P(tails) = 0.5. Each flip is independent.
- A fair 6-sided die has equal 1/6 probability for each face.
- Independent events don't remember past outcomes: the gambler's fallacy is wrong.
- Streaks are inevitable in random sequences; humans systematically underestimate this.
- The law of large numbers says that average outcomes approach expected values over many trials, but not necessarily over a few.
Coin Flips: The Basics
A coin flip has two outcomes: heads (H) or tails (T). On a fair coin, each is 50% likely.
Single flip:
- P(H) = 0.5
- P(T) = 0.5
- P(H or T) = 1.0 (something must happen)
Two flips (independent):
- P(HH) = 0.5 × 0.5 = 0.25
- P(HT) = 0.25
- P(TH) = 0.25
- P(TT) = 0.25
- Total = 1.0
Each sequence of two flips is equally likely. There's no "natural" outcome.
Multiple flips:
P(exactly k heads in n flips) = C(n, k) × 0.5^n
where C(n, k) is "n choose k." This is the binomial distribution.
| n flips | P(exactly half heads) |
|---|---|
| 2 | 50% |
| 10 | 24.6% |
| 100 | 7.96% |
| 1,000 | 2.52% |
Notice: getting exactly half heads becomes less likely as flips increase, even though the average is still 50%. The distribution flattens, so "near 50%" remains likely but "exactly 50%" doesn't.
The Streak Problem
If you flip a fair coin 10 times, what's the chance of seeing at least one streak of 4 consecutive heads or tails?
About 46%.
Most people guess much lower. Random sequences contain streaks much more often than intuition suggests. This is why "feeling hot" or "due for a turnaround" are illusions: long sequences of identical outcomes are normal, not anomalous.
In 100 flips, the probability of at least one streak of 6 in a row exceeds 80%. In a year of daily coin flips (365 trials), getting at least one streak of 8 is virtually certain.
The takeaway: humans see patterns in randomness because random data has more patterns than we expect.
Dice Probability
A standard 6-sided die has six equally likely outcomes:
- P(any specific face) = 1/6 ≈ 16.67%
- P(even number) = P(2, 4, or 6) = 3/6 = 50%
- P(number > 4) = P(5 or 6) = 2/6 ≈ 33.3%
Two dice (sum):
When rolling two dice, the sum ranges from 2 to 12, but the sums are not equally likely. There are more combinations that produce a 7 than any other sum.
| Sum | Combinations | Probability |
|---|---|---|
| 2 | 1 (1+1) | 1/36 ≈ 2.78% |
| 3 | 2 | 2/36 ≈ 5.56% |
| 4 | 3 | 3/36 ≈ 8.33% |
| 5 | 4 | 4/36 ≈ 11.11% |
| 6 | 5 | 5/36 ≈ 13.89% |
| 7 | 6 | 6/36 ≈ 16.67% |
| 8 | 5 | 5/36 ≈ 13.89% |
| 9 | 4 | 4/36 ≈ 11.11% |
| 10 | 3 | 3/36 ≈ 8.33% |
| 11 | 2 | 2/36 ≈ 5.56% |
| 12 | 1 | 1/36 ≈ 2.78% |
7 is the most common sum, which is why so many casino games and board games center on it. The distribution is a perfect triangle.
Expected Value of a Die Roll
The expected value of a single die roll:
E = (1 + 2 + 3 + 4 + 5 + 6) / 6 = 21 / 6 = 3.5
For two dice: 7 (twice the single-die expected value).
Expected value is the long-run average. After a few rolls you might see averages of 2 or 5, but after 10,000 rolls, the average will be very close to 3.5.
The Law of Large Numbers
The law of large numbers states: as the number of trials grows, the average outcome converges to the expected value.
Coin flip example:
- 10 flips: might see 7 heads (70%)
- 100 flips: probably 45–55 heads (45–55%)
- 1,000 flips: probably 470–530 heads (47–53%)
- 10,000 flips: probably 4,900–5,100 heads (49–51%)
The law of large numbers does NOT say that past outcomes "balance out." If you have 8 heads in 10 flips, future flips don't try to compensate. The coin has no memory.
Instead, the law says that the past 8 heads become a smaller fraction of the total as new flips arrive. After 1,000 flips, those 8 heads are negligible relative to the 500ish heads from new flips.
The Gambler's Fallacy
The gambler's fallacy is the belief that past outcomes affect future probabilities of independent events. "It's been 7 reds in a row at the roulette wheel, so black is due."
This is wrong. The roulette wheel has no memory. The next spin is still about 47.4% red, 47.4% black, 5.3% green (American roulette).
The mirror image is the "hot hand" fallacy: believing recent successes increase the chance of continued success. In sports, this has some genuine support due to psychological factors and skill. In pure random processes like dice and roulette, it has none.
Combining Probabilities
Multiple coins: Probabilities of all possible sequences multiply.
P(5 heads in a row) = 0.5⁵ = 1/32 ≈ 3.13% P(10 heads in a row) = 0.5¹⁰ ≈ 0.098%
Compound events: Use addition for "or" (mutually exclusive) and multiplication for "and" (independent).
P(roll a 6 with at least one of two dice) = 1 − P(no 6 on either) = 1 − (5/6)² = 1 − 25/36 = 11/36 ≈ 30.6%
(Use complement counting: "at least one" is easier as 1 minus "none.")
Worked Examples
Example 1: Best of 7 series. Two evenly matched teams. P(team wins exactly 4 games to win series) is the binomial probability of getting 4 successes in 4–7 trials. The total probability of winning the series for either team is 50%, but the distribution of how many games it takes is:
- 4 games: 12.5%
- 5 games: 25%
- 6 games: 31.25%
- 7 games: 31.25%
A "sweep" is the least common outcome despite being most decisive.
Example 2: Birthday paradox. With 23 people in a room, the probability that at least two share a birthday is about 50%. With 50 people, over 97%. The math: P(no match in n people) = 365 × 364 × ... × (365 − n + 1) / 365ⁿ. The result feels impossibly high because intuition compares one person to one specific date, not all pairs of people to all dates.
Example 3: Lottery odds. Picking 6 numbers from 49 (typical lottery format): C(49, 6) = 13,983,816 combinations. P(winning) = 1 / 13,983,816 ≈ 7.15 × 10⁻⁸. Buying 100 tickets raises your odds to 100 / 13,983,816, still effectively zero.
Common Mistakes
Gambler's fallacy. Independent events have no memory.
Underestimating streak frequency. Long random sequences have streaks much more often than intuition suggests.
Confusing P(specific sequence) with P(matching outcomes). Getting HHHHHHHHHH in 10 flips and getting HTTHTTHHTH have the same exact probability (1/1024). The former feels weirder because we have more mental patterns for "alternating" than for "all heads."
Ignoring the multiplication rule. P(A and B) = P(A) × P(B) only for independent events. Drawing cards without replacement is dependent.
Treating expected value as a guarantee. Expected value is a long-run average, not a prediction of any single trial.
Confusing probability with odds. A 25% probability is 1:3 odds. They describe the same event differently.
Practical Scenarios
Scenario 1: Fair game design. A board game designer wants players to have ~50% chance of moving forward each turn. A coin flip (or a die roll where 4–6 counts as success) achieves this exactly.
Scenario 2: Deciding fairly. Three friends need to pick one to go first. Roll a die: 1–2 → A, 3–4 → B, 5–6 → C. Each has 33.3%.
Scenario 3: Lottery-style giveaway. Pick a random number 1–1,000. Anyone matching wins. With 1,000 entrants picking independently, expected unique matches: 1,000 × 1/1,000 = 1. Expected number of zero-match scenarios: about 37% (e^−1).
Scenario 4: A/B test design. Need to detect a 5% conversion lift with 80% statistical power. Sample sizes scale with variance and effect size. The math behind these calculators is just probability applied to large samples.
FAQ
What is the probability of getting heads on a coin flip? On a fair coin, exactly 50% (0.5). The probability of getting a specific sequence over multiple flips multiplies: P(HH) = 0.25, P(HHH) = 0.125, etc.
Are coin flips truly random? Physical coin flips are deterministic in principle (governed by physics) but chaotic enough that the outcome is unpredictable in practice. Digital coin flip simulators use pseudorandom number generators that are statistically random.
What is the most common sum when rolling two dice?
- There are 6 ways to roll a 7 (1+6, 2+5, 3+4, 4+3, 5+2, 6+1) out of 36 total combinations, or 1/6 of all rolls.
Is the gambler's fallacy real? The fallacy itself is real; people believe it. The mathematical belief behind it is false. Independent random events don't depend on prior outcomes.
Why do streaks happen in random sequences? Because randomness produces more streaks than intuition expects. In 100 fair coin flips, you'll almost certainly see at least one streak of 6 or more in a row.
What is expected value? The long-run average outcome of a probabilistic process. For a fair die, expected value of a single roll is 3.5 (average of 1–6).
How can I make a "random" choice if I don't have a coin? Use any process you trust to be unbiased: a random number generator, a stranger's footstep count, a digital coin flip simulator. Modern phones have built-in random tools.
Related Tools
The Coin Flip Simulator and Dice Roller let you experiment with the probabilities described here. The Probability Calculator handles general probability questions, and the Random Number Generator provides uniformly random integers over any range.
Related Articles
Final Thoughts
Coin flips and dice rolls are the cleanest demonstrations of probability, and the clearest illustrations of how badly human intuition handles randomness. We see patterns where there are none, expect compensation where the process has no memory, and underestimate how often unlikely sequences will appear. Internalize a few facts (independence, streaks, expected value, the law of large numbers) and you'll be more reliable than most professionals in handling probabilistic questions. The math is small; the cognitive correction is large.