The Miracle Sign
One bad night, one lucky fix. After his worst Tuesday of the year, the bartender puts up a sign and tips double. Dr. Bayes suggests a less miraculous explanation.
Behind the joke.
An unusually bad night tends to be followed by a more ordinary one, whether or not anything changes. That pull back toward the average is regression to the mean, and it makes any fix applied right after a low point look effective. Without a comparison, such as nights without the sign, the doubling tells us little about the sign. The bartender then turns the same logic on Dr. Bayes's Saab, with irritating accuracy.
The transcript.
For readers who prefer dialogue without zooming in.
Read the dialogue ↘
Bartender: Worst tips I've had all year on Tuesday.
Dr. Bayes: Tragic.
Bartender: So I put up the sign.
Bartender: Tips doubled the next night.
Dr. Bayes: And therefore?
Bartender: The sign works.
Dr. Bayes: Or Tuesday was unusually bad, and Wednesday simply moved back toward normal.
Bartender: Regression to the mean?
Dr. Bayes: Exactly.
Bartender: Interesting. Your Saab started right after the mechanic adjusted it yesterday.
Bartender: Regression to the mean?
Dr. Bayes: I dislike how quickly you're learning.





