Introduction to the Binomial Distribution - AP Statistics

SECTION 1

Historical Context & Motivation

Long before statisticians formalized hypothesis testing or confidence intervals, mathematicians grappled with a deceptively simple question: if the probability of a particular outcome on any single trial is known, how can we predict the number of times that outcome will occur across many trials? This problem arose naturally in gambling, insurance, and early scientific inquiry, and its resolution gave birth to one of the most important discrete probability distributions in all of statistics — the binomial distribution. Understanding its historical roots helps illuminate why the binomial model remains central to probability theory and why it appears so frequently on the AP Statistics exam.

1654 • Pascal & Fermat Correspondence

Blaise Pascal and Pierre de Fermat exchanged letters about the problem of points — how to fairly divide stakes in an interrupted game of chance. Their analysis required counting favorable outcomes across repeated trials, laying the conceptual groundwork for binomial counting.

1713 • Bernoulli's Ars Conjectandi

Jacob Bernoulli's posthumous masterpiece formally introduced Bernoulli trials — independent experiments with exactly two outcomes — and proved the law of large numbers, connecting the binomial distribution to long-run relative frequencies.

1733 • De Moivre's Normal Approximation

Abraham de Moivre discovered that as the number of trials grows, the binomial distribution can be closely approximated by a bell-shaped curve, foreshadowing the normal distribution and the Central Limit Theorem.

1809 • Laplace's Théorie Analytique

Pierre-Simon Laplace synthesized combinatorial probability into a comprehensive framework, rigorously deriving the binomial probability formula using the binomial theorem from algebra and extending its applications to demography and astronomical error analysis.

20th C. • Modern Statistical Practice

The binomial distribution became a cornerstone of quality control (accept/reject sampling), clinical trials (success/failure outcomes), and survey methodology. It remains one of the first distributions taught in every introductory statistics course.

The central question the binomial distribution addresses is straightforward yet powerful: given n independent trials, each with the same probability p of success, what is the probability of obtaining exactly k successes? Answering this question requires combining the algebra of combinations with the multiplication rule for independent events, and doing so yields a formula of remarkable elegance and broad applicability.

SECTION 2

Core Principles & the BINS Conditions

Before applying the binomial model, you must verify that the experimental setup satisfies four conditions often remembered by the acronym BINS. These conditions are not mere formalities; violating any one of them invalidates the entire probability model and can lead to seriously incorrect conclusions. The AP exam regularly tests whether students can correctly identify binomial settings and explain why a given scenario does or does not qualify.

Binary Outcomes

Each trial must result in one of exactly two outcomes: success (the outcome of interest) or failure. These labels are arbitrary — 'success' need not be desirable. For example, success might be 'the patient experiences a side effect.'

Independent Trials

The outcome of any one trial must not influence the outcome of any other trial. In practice, sampling without replacement from a finite population violates strict independence, but the 10% condition (sample size ≤ 10% of population) makes the dependence negligible.

Number of Trials is Fixed

The value n must be determined in advance. If you keep going until you reach a certain number of successes (rather than a certain number of trials), you have a geometric or negative binomial setting, not a binomial one.

Same Probability of Success

The probability p must remain constant from trial to trial. If external conditions change — for example, a basketball player becomes fatigued — the constant-probability assumption may break down, and a different model is needed.

✦

KEY TAKEAWAY

Think of a binomial experiment like a factory assembly line that stamps identical coins: the line runs for exactly n stamps (fixed number), each coin is stamped independently (one coin's result doesn't jam the machine for the next), every stamp is either 'acceptable' or 'defective' (binary), and the machine's defect rate p stays constant throughout the run. Change any of these conditions — say the machine overheats and gets worse over time — and the binomial model no longer fits.

When all four BINS conditions are satisfied, we say the random variable X — the count of successes — follows a binomial distribution with parameters n and p, written compactly as X ~ B(n, p). The possible values of X are the integers 0, 1, 2, …, n.

SECTION 3

Visualizing the Binomial Distribution

A probability histogram is the standard way to visualize a discrete distribution. The following diagram shows the binomial probability distribution for n = 10 and p = 0.5, where each bar's height represents P(X = k). Notice how the distribution is perfectly symmetric when p = 0.5, centered at np = 5, and tapers off toward the extremes.

Probability histogram for X ~ B(10, 0.5). The dashed pink line marks the mean μ = np = 5. When p = 0.5, the distribution is perfectly symmetric about the mean.

Several features of this histogram deserve emphasis. First, the tallest bar appears at k = 5, confirming that the most probable outcome is five successes in ten trials when each trial is equally likely to succeed or fail. Second, the probabilities decrease symmetrically as k moves away from the center — obtaining 0 or 10 successes is extremely rare (about 0.1%). Third, the sum of all bar heights equals 1, as required by any legitimate probability distribution. When p ≠ 0.5, the distribution becomes skewed — right-skewed when p < 0.5 and left-skewed when p > 0.5.

SECTION 4

The Mathematical Framework

The binomial probability formula emerges from combining two ideas: the multiplication rule for independent events (to compute the probability of one specific sequence of successes and failures) and combinatorics (to count how many such sequences exist). The derivation is both accessible and instructive at the AP level.

Deriving the Binomial Probability Formula

Consider n independent Bernoulli trials, each with success probability p and failure probability q = 1 − p. For any specific sequence containing exactly k successes and (n − k) failures, the probability is p^k × q^n−k (by the multiplication rule for independent events). However, those k successes can occur in any of C(n, k) arrangements, and since these arrangements are mutually exclusive, we sum to obtain the total probability.

BINOMIAL PROBABILITY FORMULA

P(X = k) = C(n, k) × p^k × (1 − p)^(n − k)

where C(n, k) = n! / [k! × (n − k)!] is the binomial coefficient (read 'n choose k'), n = number of trials, k = number of successes, p = probability of success on a single trial, and 0 ≤ k ≤ n.

MEAN (EXPECTED VALUE)

μ_X = n × p

The mean of a binomial random variable equals the number of trials times the probability of success per trial. For n = 20 and p = 0.3, we expect μ = 6 successes on average.

STANDARD DEVIATION

σ_X = √(n × p × (1 − p))

The standard deviation measures the typical spread of the count of successes around the mean. Variance is σ² = n × p × (1 − p). Note that σ is maximized when p = 0.5 and decreases as p approaches 0 or 1.

💡

Calculator Tip

On the TI-83/84, use binompdf(n, p, k) to compute P(X = k) and binomcdf(n, p, k) to compute P(X ≤ k). These are found under 2nd → DISTR. On the AP exam, you should name the distribution and the parameters in your work, then report the calculator output.

SECTION 5

How Parameters Shape the Distribution

The shape of a binomial distribution depends critically on both n and p. Understanding this relationship is essential for interpreting binomial probabilities and for anticipating when the normal approximation becomes appropriate. The diagram below compares three binomial distributions that share the same n = 10 but have different values of p.

Three binomial distributions with n = 10 but different values of p. Notice how p < 0.5 produces right skew, p = 0.5 produces symmetry, and p > 0.5 produces left skew. The distributions for p = 0.2 and p = 0.8 are mirror images of each other.

The visual relationship between p and skewness has an elegant algebraic explanation: the distribution for B(n, p) is the mirror image of B(n, 1 − p), because relabeling 'success' and 'failure' swaps p and 1 − p. This is why the histogram for p = 0.2 is the reflection of the histogram for p = 0.8 across the midpoint of the k-axis.

Parameters	Mean (μ)	Std Dev (σ)	Shape
n = 10, p = 0.2	2	≈ 1.265	Right-skewed
n = 10, p = 0.5	5	≈ 1.581	Symmetric
n = 10, p = 0.8	8	≈ 1.265	Left-skewed
n = 50, p = 0.5	25	≈ 3.536	Approximately normal

How n and p affect the mean, spread, and shape of the binomial distribution

SECTION 6

Worked Example

A pharmaceutical company reports that 70% of patients who take a new allergy medication experience symptom relief. A clinic administers the medication to 15 randomly selected patients. Find the probability that exactly 12 patients experience relief, then find the mean and standard deviation of the number of patients who experience relief.

Binomial Probability Calculation

1
Step 1 — Verify the BINS Conditions

Binary: Each patient either experiences relief (success) or does not (failure). Independent: The patients are randomly selected, and whether one patient experiences relief does not affect another (assuming the patient population is large enough that the 10% condition is met). Number fixed: n = 15 patients is fixed in advance. Same probability: p = 0.70 for each patient. All four conditions are satisfied, so X ~ B(15, 0.70).
2
Step 2 — Identify Parameters and Goal

We have n = 15, p = 0.70, and we want P(X = 12). We also need μ and σ.
3
Step 3 — Compute the Binomial Coefficient

C(15, 12) = 15! / (12! × 3!) = (15 × 14 × 13) / (3 × 2 × 1) = 2730 / 6 = 455.

C(15, 12) = 455
4
Step 4 — Apply the Binomial Formula

P(X = 12) = C(15, 12) × (0.70)¹² × (0.30)³ = 455 × (0.70)¹² × (0.30)³. Now (0.70)¹² ≈ 0.013841 and (0.30)³ = 0.027. So P(X = 12) = 455 × 0.013841 × 0.027 ≈ 455 × 0.000374 ≈ 0.1700.

P(X = 12) ≈ 0.1700
5
Step 5 — Compute the Mean and Standard Deviation

μ = np = 15 × 0.70 = 10.5. The standard deviation is σ = √(np(1 − p)) = √(15 × 0.70 × 0.30) = √3.15 ≈ 1.775.

μ = 10.5 patients; σ ≈ 1.775 patients
6
Step 6 — Interpret in Context

There is approximately a 17.0% probability that exactly 12 out of 15 patients will experience symptom relief. On average, about 10.5 of the 15 patients will experience relief, with a typical deviation of about 1.8 patients from that average.

ℹ

Verification

On a TI-84: binompdf(15, 0.70, 12) returns approximately 0.1700, confirming our hand calculation. Always verify your work with the calculator on the AP exam, but show your setup (distribution name, parameters, and the probability statement) to earn full credit.

SECTION 7

Strengths & Limitations of the Binomial Model

The binomial distribution is one of the most widely applicable models in statistics, but like any model, it rests on assumptions that may or may not hold in practice. Recognizing when those assumptions are reasonable — and when they break down — is just as important as knowing how to calculate binomial probabilities. The AP exam frequently presents scenarios that look binomial at first glance but subtly violate one of the BINS conditions.

Strengths	Limitations
Simple two-parameter model (n and p) that is easy to specify and communicate.	Requires strict independence; correlated trials (e.g., disease spread through a household) violate this assumption.
Exact probabilities are available via the formula or calculator commands (binompdf, binomcdf).	Sampling without replacement from a small population changes p from trial to trial; use the hypergeometric distribution instead or verify the 10% condition.
Well-characterized mean (np) and standard deviation (√(npq)) allow quick summary statistics.	Only models count data with a fixed ceiling n; if the number of trials is not fixed, consider geometric or negative binomial models.
Normal approximation extends its usefulness to large n, simplifying cumulative calculations.	Assumes constant probability p; situations where conditions change (learning effects, fatigue) require more complex models.

Strengths and limitations of the binomial distribution

✦

KEY TAKEAWAY

Think of the binomial model as a perfectly calibrated scale: it gives you exact, reliable measurements as long as the object being weighed holds perfectly still (constant p) and each measurement is independent. The moment the object is shifting or the measurements start influencing each other, the scale's readings become unreliable — not because the instrument is broken, but because the assumptions it was designed for no longer hold.

SECTION 8

Connections to Other Distributions

The binomial distribution does not exist in isolation; it is part of a rich family of probability distributions, and understanding its relationships to other distributions deepens conceptual mastery and helps you select the correct model on the AP exam. Several of these connections become the basis for entire later units in AP Statistics, so building the intuition now is strategically valuable.

Related Distribution	Relationship to Binomial	When to Use Instead
Bernoulli	A binomial with n = 1. Each individual trial in a binomial experiment is a Bernoulli trial.	When analyzing a single yes/no trial (one coin flip, one patient).
Geometric	Counts the number of trials until the first success, rather than successes in a fixed number of trials.	When you want to know how many trials until a success occurs (e.g., how many calls until a sale).
Normal	When np ≥ 10 and n(1 − p) ≥ 10, the binomial is well approximated by N(np, √(np(1−p))). This is the basis of the Large Counts condition for inference.	For quick cumulative probability calculations with large n, or for constructing confidence intervals for proportions.
Hypergeometric	Models successes when sampling without replacement from a finite population (p changes each trial).	When the 10% condition fails — sampling more than 10% of a finite population without replacement.

The binomial distribution in the broader family of discrete distributions

Looking ahead in the AP Statistics curriculum, the binomial distribution plays a pivotal role in inference for proportions. When you construct a confidence interval or perform a hypothesis test for a population proportion, the underlying count of successes is modeled as binomial, and the Large Counts condition (np ≥ 10 and n(1 − p) ≥ 10) is precisely the condition under which the normal approximation to the binomial is sufficiently accurate. Understanding this connection will make the inference chapters feel like natural extensions of what you learn here, rather than unrelated new material.

SECTION 9

Practice Problems

Question 1: CONCEPTUAL

A quality inspector examines 20 items from a large production run and records whether each item is defective. She finds that 3% of items in the production run are typically defective. Which of the following best explains why a binomial model is appropriate for the number of defective items in her sample?

A. The expected number of defectives (0.6) is less than 1, making it a rare event suitable for binomial modeling.
B. The probability of defective (0.03) and non-defective (0.97) are both between 0 and 1, and they sum to 1.
C. There is a fixed number of independent trials (n = 20), each with the same probability of success (p = 0.03), and each trial has exactly two outcomes.
D. The values np = 0.6 and n(1 − p) = 19.4 satisfy the conditions for using the binomial probability formula.
E. Each item is replaced after inspection, ensuring the probability remains constant across all 20 trials.

Show answer (C)

The correct answer is C. Option A is incorrect because the expected number being less than 1 does not determine whether a binomial model is appropriate — it merely describes the distribution's mean. Option B is incorrect because the binomial model does not require an equal probability of success and failure (p = 0.5); it only requires that p be constant. Option C is correct because it correctly identifies all four BINS conditions: binary outcomes (defective/not), independence (the production run is large, so the 10% condition is satisfied for n = 20), fixed number of trials (n = 20), and constant probability (p = 0.03 for each item). Option D is incorrect because while the normal approximation does require np ≥ 10 and n(1−p) ≥ 10, this condition determines when a normal approximation to the binomial is valid, not whether the binomial model itself is appropriate. Option E is incorrect because replacement is not required; what matters is that the probability stays approximately constant, which the 10% condition verifies.

Question 2: BASIC CALCULATION

A fair six-sided die is rolled 8 times. Let X be the number of times a '6' appears. What is P(X = 2)?

A. 0.1042
B. 0.2605
C. 0.3241
D. 0.0465
E. 0.2936

Show answer (B)

The correct answer is B. Here X ~ B(8, 1/6). We compute P(X = 2) = C(8, 2) × (1/6)² × (5/6)⁶ = 28 × (1/36) × (15625/46656) = 28 × 0.02778 × 0.33490 ≈ 28 × 0.009303 ≈ 0.2605. Option A (0.1042) results from using C(8, 2) × (1/6)² × (5/6)⁵, which incorrectly uses exponent 5 instead of 6 on the failure term. Option C (0.3241) results from computing P(X ≤ 2) cumulatively rather than P(X = 2) exactly. Option D (0.0465) comes from forgetting the binomial coefficient entirely: (1/6)² × (5/6)⁶ without the C(8,2) multiplier. Option E (0.2936) corresponds to P(X = 1), not P(X = 2).

Question 3: INTERMEDIATE

In a certain city, 40% of registered voters support a proposed bond measure. A polling firm randomly selects 25 registered voters from the city (the city has over 50,000 registered voters). What is the probability that fewer than 8 of the 25 selected voters support the measure?

A. 0.0294
B. 0.0736
C. 0.2735
D. 0.1091
E. 0.5858

Show answer (D)

The correct answer is D. Since the city has over 50,000 registered voters and the sample size is 25, which is less than 10% of 50,000, the 10% condition is satisfied. We model X ~ B(25, 0.40). The phrase "fewer than 8" means P(X < 8) = P(X ≤ 7) = binomcdf(25, 0.40, 7) ≈ 0.1091. Option A (0.0294) corresponds to P(X ≤ 5) = binomcdf(25, 0.40, 5) ≈ 0.0294, which uses a boundary that is too low—as if the student misidentified the cutoff as "fewer than 6." Option B (0.0736) corresponds to P(X ≤ 6) = binomcdf(25, 0.40, 6) ≈ 0.0736, which uses a boundary that is too low—as if the student misidentified the cutoff as "fewer than 7." Option C (0.2735) corresponds to P(X ≤ 8) = binomcdf(25, 0.40, 8) ≈ 0.2735, which incorrectly includes P(X = 8). This is the most common error, resulting from interpreting "fewer than 8" as "at most 8" rather than "at most 7." Option E (0.5858) corresponds to P(X ≤ 10) = binomcdf(25, 0.40, 10) ≈ 0.5858, which uses a boundary that is too high—as if the student misidentified the cutoff as "fewer than 11." The key distinction is that "fewer than 8" means X ≤ 7, not X ≤ 8.

Question 4: APPLIED

4 pts

A semiconductor manufacturer knows that 8% of a certain type of chip fails quality testing. In a batch of 30 chips, the production line is shut down for recalibration if 5 or more chips fail. (a) Define the random variable and verify that a binomial model is appropriate. (b) Calculate the probability that the production line will be shut down (i.e., 5 or more chips fail). (c) Find the mean and standard deviation of the number of failing chips. (d) If the failure rate increased to 15%, how would the probability of shutdown change? Explain without performing a full calculation.

Show answer

Part (a) — 1 point: Let X = the number of chips that fail quality testing out of 30 chips. Binary: each chip either fails (success) or passes (failure). Independent: one chip's test result does not affect another's (assuming a large production run satisfies the 10% condition). Fixed n: 30 chips. Constant p = 0.08. Therefore X ~ B(30, 0.08). Part (b) — 1 point: P(X ≥ 5) = 1 − P(X ≤ 4) = 1 − binomcdf(30, 0.08, 4) ≈ 1 − 0.9652 ≈ 0.0348. There is approximately a 3.5% chance the line will be shut down. Part (c) — 1 point: μ = np = 30 × 0.08 = 2.4 chips. σ = √(npq) = √(30 × 0.08 × 0.92) = √(2.208) ≈ 1.486 chips. Part (d) — 1 point: If p increased to 0.15, the mean number of failures would increase from 2.4 to np = 30 × 0.15 = 4.5, shifting the entire distribution to the right. Because the center of the distribution moves closer to the shutdown threshold of 5, and the standard deviation also increases (√(30 × 0.15 × 0.85) ≈ 1.96), substantially more probability mass falls at or above 5. Therefore, the probability of shutdown would increase significantly — likely to somewhere around 0.25 or higher — making recalibration far more frequent.

Question 5: CRITICAL THINKING

4 pts

A student claims that the following scenario can be modeled with a binomial distribution: 'A jar contains 12 red marbles and 8 blue marbles. You draw 10 marbles one at a time without replacement and count the number of red marbles drawn.' (a) Explain why this scenario does not strictly satisfy the conditions for a binomial distribution. (b) State a specific numerical condition that, if satisfied, would justify using the binomial as a reasonable approximation, and explain why it works. (c) Describe an alternative probability distribution that would provide an exact model for this scenario, and identify its parameters. (d) Suppose instead the jar contained 1,200 red and 800 blue marbles, and you still drew 10. Argue whether the binomial approximation is now reasonable, and if so, identify the appropriate binomial parameters.

Show answer

Part (a) — 1 point: The scenario violates the independence condition (and equivalently the constant probability condition). When sampling without replacement from a finite population, the probability of drawing a red marble changes from trial to trial. For example, if the first draw is red, P(red on 2nd draw) = 11/19, but if the first draw is blue, P(red on 2nd draw) = 12/19. Since the probability of success is not constant and the trials are not independent, a strict binomial model is inappropriate. Part (b) — 1 point: The 10% condition states that if the sample size n is no more than 10% of the population size N, the binomial is a reasonable approximation even when sampling without replacement. Here n = 10 and N = 20, so n/N = 50%, which far exceeds 10%. The condition works because when the sample is a tiny fraction of the population, removing one item barely changes the composition, making the probability of success nearly constant and the draws nearly independent. Part (c) — 1 point: The hypergeometric distribution provides an exact model. Its parameters are: N = 20 (total population), K = 12 (total red marbles, i.e., 'successes' in the population), and n = 10 (number of draws). It directly accounts for the changing probabilities caused by sampling without replacement. Part (d) — 1 point: With N = 2,000 and n = 10, we have n/N = 10/2000 = 0.5%, which is well below the 10% threshold. Removing 10 marbles from a jar of 2,000 changes the composition negligibly, so the probability of drawing a red marble stays very close to 1200/2000 = 0.60 on every draw. The binomial approximation B(10, 0.60) is therefore reasonable, where n = 10 and p = 0.60.

SUMMARY

Summary

The binomial distribution models the count of successes X in n independent trials, each with the same probability p of success. The four BINS conditions — Binary outcomes, Independent trials, fixed Number of trials, and Same probability — must be verified before applying the model. The probability formula P(X = k) = C(n, k) × p^k × (1 − p)^n−k combines the binomial coefficient with the multiplication rule for independent events.

Key summary statistics include the mean μ = np and the standard deviation σ = √(np(1 − p)). The distribution's shape depends on p: symmetric when p = 0.5, right-skewed when p < 0.5, and left-skewed when p > 0.5. When sampling without replacement, the 10% condition (n ≤ 0.10N) justifies treating the trials as approximately independent. On the calculator, use binompdf for exact probabilities and binomcdf for cumulative probabilities. The binomial distribution is foundational for later inference on proportions, where the normal approximation (valid when np ≥ 10 and n(1 − p) ≥ 10) becomes the workhorse of confidence intervals and hypothesis tests.