Poisson Distribution: Formula, Mean Rate λ and the Shape of the Curve
The Poisson distribution gives the probability of exactly k events in an interval when they happen independently at a mean rate λ. Slide λ and watch the curve change shape.
This concept builds on Normal Distribution — The Bell Curve Explained, Binomial Distribution — Counting Coin Flips. Consider exploring them first for the best experience.
Rare Events & the Rate λ
The Poisson distribution gives the probability of seeing exactly k events in one interval when events happen independently at a steady average rate λ. Imagine counting emails arriving in an hour, radioactive decays per minute or typos per page. Watch the bars grow in at λ = 1: nearly all the height sits on k = 0 and k = 1. Each stage waits for you — press "Next stage" when you are ready.
Worked Example: P(X = 2) at λ = 3
The formula is P(X = k) = λ^k × e^−λ / k!. This stage glides λ to 3 and pins the k = 2 bar with a yellow marker. Plug in: 3² = 9, e^−3 ≈ 0.0498 and 2! = 2, so P(X = 2) = 9 × 0.0498 / 2 ≈ 0.224, and the "P(X = k)" card reads 0.2240. The k = 3 bar next to it is exactly as tall, because 3³/3! = 3²/2! = 4.5.
Mean and Variance Both Equal λ
Press "Next stage" to glide λ to 5. The mean E[X] and the variance Var(X) both equal λ, so the dashed orange μ line sits at 5 and the standard deviation is √5 ≈ 2.24. Press "Simulate 100 intervals" a few times: the yellow observed dashes settle onto the bars and the "Simulated mean" card drifts toward 5.
The Mode: the Highlighted Bar
The purple highlighted bar marks the mode, floor(λ), the most likely count. At λ = 5 it is the k = 5 bar, and the k = 4 bar beside it is exactly as tall: whenever λ is a whole number, both λ and λ − 1 are modes. Move the "Rate λ" slider to 5.5 and the highlight stays on k = 5; at 6 it jumps to k = 6.
Convergence to Normal
Press "Next stage" and λ climbs from 5 to 15. Follow the purple highlighted bar: it slides right with the peak while the bars spread out and flatten, and from about λ = 8 a pink normal curve fades in over them. By λ = 15 it hugs the bar tops, because a Poisson count is a sum of many small independent counts and the CLT applies. Tick "Show normal approximation" at λ = 2 to see the poor fit at small λ.
The Binomial Limit & Real-World Uses
The last stage brings λ back to 2. Slide "Binomial trials n" to overlay Binomial(n, p = λ/n) as blue circles: at n = 5 they miss the bars, by n = 500 they sit on top of them. That is why Poisson models calls at a call centre, photons hitting a detector or defects on a chip — many chances, each rare, at a steady rate.
How It Works — Step by Step
-
1
Rare Events & the Rate λ
The Poisson distribution gives the probability of seeing exactly k events in one interval when events happen independently at a steady average rate λ. Imagine counting emails arriving in an hour, radioactive decays per minute or typos per page. Watch the bars grow in at λ = 1: nearly all the height sits on k = 0 and k = 1. Each stage waits for you — press "Next stage" when you are ready.
-
2
Worked Example: P(X = 2) at λ = 3
The formula is P(X = k) = λ^k × e^−λ / k!. This stage glides λ to 3 and pins the k = 2 bar with a yellow marker. Plug in: 3² = 9, e^−3 ≈ 0.0498 and 2! = 2, so P(X = 2) = 9 × 0.0498 / 2 ≈ 0.224, and the "P(X = k)" card reads 0.2240. The k = 3 bar next to it is exactly as tall, because 3³/3! = 3²/2! = 4.5.
-
3
Mean and Variance Both Equal λ
Press "Next stage" to glide λ to 5. The mean E[X] and the variance Var(X) both equal λ, so the dashed orange μ line sits at 5 and the standard deviation is √5 ≈ 2.24. Press "Simulate 100 intervals" a few times: the yellow observed dashes settle onto the bars and the "Simulated mean" card drifts toward 5.
-
4
The Mode: the Highlighted Bar
The purple highlighted bar marks the mode, floor(λ), the most likely count. At λ = 5 it is the k = 5 bar, and the k = 4 bar beside it is exactly as tall: whenever λ is a whole number, both λ and λ − 1 are modes. Move the "Rate λ" slider to 5.5 and the highlight stays on k = 5; at 6 it jumps to k = 6.
-
5
Convergence to Normal
Press "Next stage" and λ climbs from 5 to 15. Follow the purple highlighted bar: it slides right with the peak while the bars spread out and flatten, and from about λ = 8 a pink normal curve fades in over them. By λ = 15 it hugs the bar tops, because a Poisson count is a sum of many small independent counts and the CLT applies. Tick "Show normal approximation" at λ = 2 to see the poor fit at small λ.
-
6
The Binomial Limit & Real-World Uses
The last stage brings λ back to 2. Slide "Binomial trials n" to overlay Binomial(n, p = λ/n) as blue circles: at n = 5 they miss the bars, by n = 500 they sit on top of them. That is why Poisson models calls at a call centre, photons hitting a detector or defects on a chip — many chances, each rare, at a steady rate.
Key Takeaways
P(k) = (λᵏ × e⁻λ) / k!
Use this formula whenever you have independent rare events occurring at a constant average rate λ. It gives the exact probability of exactly k occurrences in one interval.
Mean = Variance = λ
Both the expected value and the variance equal λ. This unique property makes the Poisson easy to parameterise: a single number λ describes both the centre and the spread.
As λ Grows, Shape Approaches Normal
For λ ≥ 10, the normal approximation N(μ=λ, σ²=λ) is accurate. This lets you use z-scores and normal tables for large-λ Poisson problems.
Poisson Is the Limit of Binomial
When n is large and p is small with np = λ, Binomial(n, p) ≈ Poisson(λ). This makes Poisson the natural model for rare-event counts in large populations.
Connected Concepts
Binomial Distribution
Poisson is a limiting case of the binomial: as n → ∞ and p → 0 with np = λ fixed, binomial converges to Poisson.
Normal Distribution
For large λ, the Poisson distribution looks almost exactly like a bell curve — the CLT in action.
Central Limit Theorem
As λ grows, the standardised Poisson converges to the standard normal by the CLT.
Bayes' Theorem
Bayesian inference uses the Poisson likelihood to update beliefs about arrival rates λ.
Law of Large Numbers
Simulate more and more intervals and the observed average count settles toward λ — the law of large numbers at work.
Updated
- — The title and summary now name the Poisson distribution, its formula and mean rate λ; step 1 opens with a definition.
- — Each stage now waits for "Next stage"; new worked example P(X = 2) at λ = 3 ≈ 0.224; steps follow the highlighted bar