Survivorship Bias Explained: Examples and WWII Planes

Quick answer: Survivorship bias is a type of selection bias that occurs when we study only the people, objects or results that survived a selection process and ignore those that disappeared or failed. It can make success look more common, or risks look smaller, than they really are. During World War II, mathematician Abraham Wald studied damage on aircraft that returned: holes on returning planes showed which hits were survivable, while planes lost in combat were missing from the evidence. Before trusting successful examples, ask what happened to the ones you cannot see.

Facts checked November 22, 2026.

Look at the bombers that came home and you will see bullet holes everywhere except a few places. Where should you add armor? The obvious answer, where the holes are, is exactly wrong. This puzzle, tied to the mathematician Abraham Wald, is the most famous example of survivorship bias, and the reasoning behind it is a short lesson in conditional probability.

The Bomber Story

During World War II, Abraham Wald worked with the Statistical Research Group, a team of mathematicians at Columbia University helping the US military. Analysts could only inspect aircraft that returned from missions. Wald’s insight: the sample was filtered. Planes hit in vulnerable places, such as the engines, often never came back, so their damage was missing from the data. Few holes in an area did not mean that area was rarely hit; it meant planes hit there rarely survived. His 1943 memoranda, reprinted by the Center for Naval Analyses in 1980, developed statistical methods to estimate aircraft vulnerability from the damage on survivors; they are more technical than the dramatic retellings suggest. (The popular image of a plane covered in red dots is a modern illustration, not an original wartime damage map.)

A Worked Example With Numbers

This is a hypothetical teaching model, not Wald’s data. Suppose 100 bombers each take exactly one hit, assigned at random to one of four equally likely areas, so about 25 hits per area. The survival chances below are invented for illustration:

Area hitPlanes hitChance of returningPlanes that return with a hole there
Engine2520%5
Fuselage2590%22.5
Wings2585%21.25
Tail2595%23.75

About 72.5 planes return. Among them, only \(5 \div 72.5 \approx 7\%\) have engine holes, even though engines took 25% of all hits. In probability language, \(P(\text{engine hit} \mid \text{returned})\) is much smaller than \(P(\text{engine hit})\) because engine hits make returning unlikely. By Bayes’ rule,

\[ P(\text{engine} \mid \text{returned}) = \frac{P(\text{returned} \mid \text{engine}) \, P(\text{engine})}{P(\text{returned})} = \frac{0.2 \times 0.25}{0.725} \approx 0.069. \]

Engines receive 25% of all hits but only about 6.9% of the hits seen on returning aircraft. Someone looking only at survivors might conclude engines are rarely hit. The counting is not wrong; the sample systematically excludes the critical failures. The missing holes are the clue: armor the engines. See how to calculate probability for the basics.

Survivorship Bias Everywhere

What you seeWhat is missingThe mistake
Famous college dropouts who became billionairesThe many dropouts who did not“Dropping out leads to success”
Investment funds with great 10-year recordsFunds that closed or merged after doing badlyOverestimating average fund performance
Beautiful old buildings still standingOld buildings that fell down or were demolished“They built better in the past”
Successful startup founders’ habitsFailed founders with the same habitsCopying traits that do not cause success
Cats treated by vets after falls (a 1987 study of 132 cats at one New York hospital)Cats that died and were never brought inGeneralizing treated cats’ survival to all falls

Survivorship Bias vs Other Biases

BiasWhat goes wrong
Selection biasThe sample is not representative (survivorship bias is one type)
Survivorship biasFailures or losses vanish from the data
Confirmation biasWe favor evidence that supports what we already believe

The core issue is the denominator: 10 successful founders who dropped out of college cannot tell you a success rate unless you know how many dropouts tried the same path. Statisticians call this data “missing not at random”: whether a case disappears depends on the very outcome you care about, and the missing failures may be impossible to reconstruct without extra records.

Survivorship Bias in AI and Data

  • Benchmarks: results that look impressive get published; failed runs and weaker settings often do not (see how to read AI benchmarks).
  • Trading strategies: backtests only of strategies that “worked,” while the many that were tried and discarded stay hidden (see can AI predict the stock market?).
  • Training data: data that survived filtering or was easy to collect can skew what an AI model learns about the world.
  • Your feed: viral posts are survivors of a huge competition; the many ordinary posts are invisible, a cousin of the friendship paradox.

How to Avoid Survivorship Bias

  1. Ask what filter produced your data, and who or what did not make it through.
  2. Look for the denominator: how many tried, not just how many succeeded.
  3. Compare survivors with non-survivors on the same trait.
  4. Be suspicious of “secrets of success” drawn only from winners.

Common Misconceptions

“Survivorship bias means survivors are lucky.” Not necessarily; it means your sample is filtered, so conclusions drawn from it can be wrong.

“More data fixes it.” More filtered data repeats the same bias; you need the missing cases.

“The red-dot plane picture is a real WWII record.” It is a modern illustration of Wald’s idea.

Watch: Survivorship Bias Explained

If you learn better by watching, these videos are worth your time.

Survivorship Bias (Eddie Woo). A teacher’s clear classroom explanation with the bomber example.
How a Mathematician Saved the American Bombers in World War II – Abraham Wald & Survivorship Bias (TJ3 History). The history of Wald’s work and the Statistical Research Group.

Check Your Understanding

Five quick questions to see how much of this has stuck. Pick an answer and press Submit; you will see a short explanation either way.

Answer each question and press Submit to see your result right away.

1. Where did Wald suggest adding armor?
Found a bug? Let us know!
2. In the example, how many planes return with engine holes?
Found a bug? Let us know!
3. How is survivorship bias related to selection bias?
Found a bug? Let us know!
4. What share of returning planes have engine holes?
Found a bug? Let us know!
5. Which is an example of survivorship bias?
Found a bug? Let us know!

Frequently Asked Questions

What is survivorship bias?

The error of drawing conclusions only from cases that survived some selection process, while ignoring those that did not, which can make conclusions badly wrong.

What is the Abraham Wald airplane story?

During World War II, Wald reasoned that bullet holes on returning bombers showed where planes could survive hits; areas with few holes, like engines, were where hits were fatal, so those needed armor.

Is the red-dot airplane picture real?

No. It is a modern illustration of Wald’s idea, not a wartime image of actual damage data.

How is survivorship bias related to probability?

It confuses P(trait | survived) with P(trait). Bayes’ rule shows how surviving changes the probabilities you observe.

What are everyday examples of survivorship bias?

Billionaire dropout stories, investment funds that survived, old buildings still standing, and viral posts, each ignoring the many cases that did not make it.

Keep Exploring

Sources and Further Reading

Related to This Article

What people say about "Survivorship Bias Explained: Examples and WWII Planes - Effortless Math"?

No one replied yet.

Leave a Reply