Card 08

🏥 📊 🤔

Hospital A has 10 babies born, 7 are boys. Hospital B has 1000 babies, 510 are boys. Which result is more surprising?

💭 How to Think About This

Hospital A: 70% boys! Hospital B: 51% boys. Hospital A seems more extreme. But which is actually more UNUSUAL? Understanding this requires grasping one of probability's most important ideas: small samples produce extreme results by chance. Large samples are more reliable.

Which result do you find more surprising or unusual?

🔒 Write your thinking above to unlock hints

Flip a fair coin 10 times—getting 7 heads (70%) isn't unusual at all! It happens about 12% of the time. But flip 1000 times and get 510 heads (51%)? That's actually quite close to the expected 500. With 1000 flips, getting 700 heads (70%) would be essentially impossible!

LAW OF LARGE NUMBERS: As sample size increases, the observed percentage gets closer to the true percentage. Small samples can deviate wildly from the truth. Large samples are more stable and reliable. This is why polls survey thousands, not tens—small samples are too noisy!

Small samples produce misleading extremes: "5 out of 5 dentists recommend..." (only asked 5!) • "This village has the highest cancer rate!" (population: 47) • "My friend tried it and it worked!" (sample size: 1). Extreme percentages from tiny samples should trigger immediate skepticism.

70% from 10 babies isn't surprising at all—it's EXPECTED randomness from a tiny sample. But 51% from 1000 babies means the true rate is probably very close to 51%. The "boring" result from the large sample is more informative! Counter-intuitively, Hospital B's result tells us more.

Hospital B's 51% is actually more surprising! 70% from 10 births is common by chance; 51% from 1000 is stable data.

Your intuition focused on the more extreme percentage, but that's exactly the trap this problem reveals.

The math intuition:

• 10 coin flips: Getting 7 heads (70%) happens ~12% of the time

• 1000 coin flips: Getting 510 heads is very close to expected 500

• With 1000 flips, 70% (700 heads) would be astronomically unlikely

The key principle:

• Small samples → high variance → extreme results common

• Large samples → low variance → results cluster near truth

The protection: Always ask "How large is the sample?" before being impressed by percentages.

You're right to think sample size matters! This is the LAW OF LARGE NUMBERS in action. Small samples are noisy—they can deviate wildly from the true average. Large samples are stable—they reliably estimate the truth. 70% from 10 vs 51% from 1000 tells very different stories.

Let's be precise: Flip a coin 10 times—getting 7+ heads happens about 17% of the time. Not unusual at all! But flip 1000 times and get 510 heads—that's well within normal range (expected: 500 ± about 15). The larger sample gives us a reliable estimate of the true rate.

This principle explains why: • Small schools top AND bottom academic rankings (small samples vary more) • "5 out of 5 dentists" claims are weak evidence • Anecdotes ("worked for my friend!") prove nothing • Medical trials need thousands of participants. Sample size determines signal vs noise!

So which is MORE SURPRISING? Hospital B's 51%! Because with 1000 babies, we'd expect close to 50%. Getting 51% is well within normal variation but represents a real signal. Hospital A's 70% is just noise from a tiny sample—interesting but meaningless. Large samples reveal truth!

Hospital B's 51% is more surprising! You correctly identified that sample size changes everything.

Your intuition about sample size effects is spot-on. Here's why it matters:

The math:

• 10 births: Standard deviation ≈ 15 percentage points—wild swings expected

• 1000 births: Standard deviation ≈ 1.5 percentage points—very stable

• 70% from 10 = normal noise; 51% from 1000 = real signal

The law of large numbers:

As samples grow, observed averages converge to true averages. Large samples are trustworthy; small samples are noisy.

Your habit to build: Before reacting to ANY statistic, ask: "How large is the sample?" This single question prevents countless reasoning errors.

You recognized that the larger sample tells us more! 1000 babies is enough to be confident the true rate is close to 51%. Hospital A's 70% could easily be random noise from just 10 births. Large samples give reliable information; small samples are noisy and misleading.

With 1000 births, we'd expect results very close to 50% if boys and girls are equally likely. Getting 51% means either: (a) slight true bias toward boys, or (b) normal statistical variation. Either way, it's real data! Hospital A's 70% tells us almost nothing—too few cases.

LAW OF LARGE NUMBERS: As sample size increases, observed percentages get closer to true percentages. With 10 births, wild variation is normal. With 1000 births, results cluster near reality. This is why serious polls use thousands of respondents—small samples are unreliable!

Your sample-size awareness protects you from: • "5 out of 5 dentists" claims (tiny sample!) • Small school rankings (they're also the worst schools!) • Anecdotal evidence ("worked for my friend!") • Early-stage trials with few participants. Always ask: "How many cases is this based on?"

Exactly right! Hospital B's 51% from 1000 babies is genuinely informative—Hospital A's 70% is just noise.

You showed excellent probabilistic thinking by recognizing that sample size determines reliability.

Why you're right:

• 70% from 10 cases = expected random variation (happens ~12% of the time)

• 51% from 1000 cases = stable estimate of true rate

• The "boring" result is more informative!

The key principle:

• Small samples → high variance → extreme results common

• Large samples → low variance → results reveal truth

Your protection: Before believing any statistic, ask: "How large is the sample?" This habit prevents countless reasoning errors.

🤔 Which thinking lens(es) did you use?

Select all the lenses you used:

Learn more about the 7 Thinking Lenses →

👨‍👩‍👧 For Parents & Teachers

🌱 A Small Everyday Story

"This tiny school has the best test scores
in the entire state!" the article proclaimed.
But next year, the same school had the worst.
With only 15 students taking the test,
a few smart or struggling kids
swing the average wildly.
Sample size was the story, not teaching quality.

See more guidance →

🧠 Thinking habits this builds:

Automatically asking "how large is the sample?" when seeing statistics
Understanding that extreme results often come from small samples
Recognizing that large samples produce more reliable estimates
Being skeptical of impressive percentages without sample size context

🌿 Behaviors you may notice (and reinforce):

"But how many people was that based on?" questions
Skepticism about statistics from very small groups
Understanding why anecdotes aren't reliable evidence
Recognizing the "small school" paradox in rankings

How to reinforce: When discussing statistics together, always ask: "What's the sample size?" Make it a reflex—percentages without sample size are incomplete information.

🔄 When ideas are still forming:

Some learners may think they need thousands of samples for ANY conclusion. Help them understand that required sample size depends on the question—sometimes 100 is enough, sometimes 10,000 isn't. The key is understanding when small samples are risky.

Helpful response: "How certain do we need to be? How different is the result from what we'd expect by chance? Those questions determine how much data we need."

🔬 If you want to go deeper:

Explore why political polls report "margin of error"
Discuss the "small school movement" and why it failed
Look up the law of large numbers and central limit theorem

Key concepts (for adults): Sample size, law of large numbers, variance, margin of error, statistical significance, central limit theorem, small sample fallacy, anecdotal evidence.

Hospital A has 10 babies born, 7 are boys. Hospital B has 1000 babies, 510 are boys. Which result is more surprising?

🔄 Compare All Perspectives

🏥 "70% seems more extreme"

🤔 "Sample size matters"

📊 "51% from 1000 is real data"

🤔 Which thinking lens(es) did you use?