Plinko Probability

What is Plinko Probability?

Plinko Probability turns the famous game-show board into a rigorous statistics laboratory. Each ball dropped through the board hits a series of pegs, bouncing left or right at each one independently. After traveling through all the rows, the ball lands in one of the bins at the bottom, and the histogram of where many balls land reveals the underlying probability distribution. When each peg has an equal 50/50 chance of sending the ball left or right, the number of right-bounces out of n rows follows a binomial distribution. The critical insight — the Central Limit Theorem in action — is that this binomial distribution increasingly resembles a normal (Gaussian, bell-shaped) curve as the number of rows grows. This happens because the final bin position is the sum of many small independent random steps, and the CLT says that sums of independent random variables tend toward a normal distribution regardless of the individual step distribution. The mean of the distribution is μ = n × p (where p is the probability of going right) and the standard deviation is σ = √(n × p × (1-p)). Plinko makes these abstract formulas physically visible: you can literally watch the bell curve assemble itself one ball at a time, and you can see how changing the bias (prob_right) skews the histogram away from the center.

Parameters explained

Number of Rows

Sets the number of peg rows the ball passes through, from 1 to 12 (default 6). Each row adds one independent random bounce. More rows produce more possible landing bins and a distribution that more closely approximates a continuous normal curve. With only 1 row there are just 2 possible outcomes (left or right); with 12 rows there are 13 possible bins and the bell shape is clearly visible after dropping enough balls.

Balls to Drop

Controls how many balls are dropped in a single simulation run, from 1 to 100 (default 10). Dropping more balls fills the histogram more completely, reducing random fluctuations and making the theoretical distribution shape easier to see. With only 10 balls the histogram is noisy and irregular; with 100 balls the bell curve emerges clearly. This directly demonstrates the law of large numbers: empirical frequencies converge to theoretical probabilities as sample size grows.

Probability Right(%)

Sets the probability that a ball bounces right at each peg, from 10% to 90% (default 50%). At 50% the board is symmetric and the distribution peaks in the center. Reducing or increasing prob_right shifts the peak toward the left or right bins respectively, producing a skewed distribution. This shows how a biased random process translates into a shifted distribution — connecting to real-world applications like biased coins, genetic inheritance, or quality control testing.

Common misconceptions

• MisconceptionPast balls influence where the next ball will land — if many balls fell right recently, the next is more likely to go left.

  CorrectEach peg encounter is fully independent of every other. There is no memory in the board. This is the gambler's fallacy: the belief that random outcomes must 'balance out' in the short run. In reality, the probability at each peg is exactly prob_right on every bounce regardless of what previous balls did. The long-run distribution converges to the expected shape not because individual outcomes compensate, but because the law of large numbers averages out over many trials.

• MisconceptionAll bins are equally likely because the ball makes random choices at every peg.

  CorrectThe choices are random, but the number of different paths leading to each bin varies enormously. The center bin can be reached by many different sequences of lefts and rights (e.g., C(6,3) = 20 paths for a 6-row board), while the extreme bins can only be reached by one path each (all right or all left). More paths to a bin means higher probability — exactly the logic of Pascal's triangle and the binomial coefficients.

• MisconceptionIncreasing the number of rows makes the distribution narrower because there are more pegs to average over.

  CorrectMore rows make the distribution wider, not narrower. Standard deviation grows as √(n × p × (1-p)), which increases with n. However, if you normalize by dividing by n (looking at the fraction of right-bounces rather than the count), the normalized distribution does become narrower as n increases. The absolute spread grows; the relative spread shrinks. This is an important distinction in understanding the CLT.

• MisconceptionThe normal distribution that emerges is exact for any number of rows.

  CorrectThe binomial distribution only approximates a normal distribution, and the approximation improves as n (number of rows) increases and as p stays near 0.5. For small n or extreme p values (near 10% or 90%), the distribution is visibly discrete and asymmetric rather than bell-shaped. The simulation makes this visible: compare the histogram at 2 rows vs. 12 rows, or at prob_right 50% vs. 20%.

How teachers use this lab

1. Law of large numbers demo: set num_rows to 6, prob_right to 50%, and drop balls progressively — first 5, then 20, then 100. After each batch, ask students to describe the histogram shape. Students observe the histogram stabilizing as num_balls increases, illustrating how empirical frequencies converge to theoretical probabilities with more trials. For a CLT demo, fix num_balls at 100 and compare num_rows at 3, 6, and 12 to show how increasing independent steps per ball produces a distribution that more closely approximates a normal curve.
2. Standard deviation scaling: keep prob_right at 50% and num_balls at 100. Run with num_rows at 3, then 6, then 12. Students calculate the theoretical σ = √(n × 0.5 × 0.5) for each and compare to the histogram spread. This concretizes the √n scaling relationship before introducing formal variance formulas.
3. Symmetry-breaking with bias: start at prob_right 50% with num_rows 6 and num_balls 100 to establish a symmetric baseline. Then shift prob_right to 30% and run again. Students compare the peak location to the theoretical mean μ = n × p = 6 × 0.3 = 1.8 (approximately bin 2). This connects binomial mean formula to a visual result.
4. Empirical vs. theoretical probability: with num_rows 6 and prob_right 50%, drop num_balls 100 and record the fraction of balls in the center bin. Compare to the theoretical probability P(3 rights) = C(6,3)/2^6 = 20/64 ≈ 31.25%. Repeat with a fresh run to show variation, reinforcing that empirical probability fluctuates around the theoretical value.
5. Gambler's fallacy discussion: drop 10 balls and note the uneven histogram. Ask students whether they expect the next ball to 'correct' the imbalance. Then drop 90 more balls and show the distribution averaging out — not because any ball compensated, but because large samples dominate small-sample noise. This directly addresses the independence misconception.