The Texas Sharpshooter Fallacy

A Texan fires randomly at a barn, then paints a bullseye around the tightest cluster of holes and claims to be a sharpshooter. This is how we fool ourselves into seeing patterns where none exist.

🎯 Be the Texas Sharpshooter

1. Fire Randomly

2. Find "Clusters"

3. Paint Targets

Shots Fired

"Bullseyes"

Click "Fire Randomly" to shoot 10 random bullets at the barn. Notice how some bullets naturally cluster together—by pure chance!

🧠 Can You Spot Real Patterns?

One of these patterns is truly random. The other has a subtle structure. Can you tell which is which? (Hint: It's harder than you think!)

Pattern A

Pattern B

🗺️ The Cancer Cluster Problem

Journalists often report "cancer clusters"—areas where cancer rates seem unusually high. But random data ALWAYS has clusters. Press the button to generate a new random distribution of cases.

Every random distribution has clusters. The question isn't "is there a cluster?" but "is this cluster MORE clustered than random chance predicts?" Without proper statistical analysis, we see patterns everywhere.

📚 Real-World Examples

🔮 Nostradamus Prophecies

Vague quatrains written in 1555 are "matched" to modern events AFTER they happen. With enough text and liberal interpretation, anything fits.

📈 Stock Market "Patterns"

Technical analysts find "head and shoulders," "cup and handle" patterns in random price movements. Studies show these patterns don't predict better than chance.

🧬 P-Hacking in Science

Test enough variables and something will be "statistically significant" (p < 0.05). If you test 20 things, expect 1 false positive by chance alone!

🎰 The Hot Hand Fallacy

Basketball players seem to go on "hot streaks"—but studies show shooting success is mostly independent. We see runs in randomness and assume meaning.

"At best, the occurrence of a cluster in the data is the basis not for a causal conclusion, but for the formation of a causal hypothesis which needs to be tested. Patterns in data can be useful for forming hypotheses, but they are not themselves sufficient evidence of a causal connection."

— The Fallacy Files

🛡️ How to Avoid This Fallacy

Pre-Register Hypotheses

Decide WHAT you're looking for BEFORE examining the data. Paint the target before you shoot!

Correct for Multiple Comparisons

If testing many hypotheses, adjust your significance threshold. The Bonferroni correction divides α by the number of tests.

Replicate Findings

A pattern found once might be chance. Confirm it in new, independent data.

Consider Base Rates

How often would this pattern appear by chance? Random data IS clustered—that's expected!