Mark has famously stated that he believes the odds of aligning 7 facial features on two random people is approximately 1 in 10 million. His math is very simple. He estimates that there is a 1 in 10 chance that one facial feature will align, thus multiplying 1/10 to the 7th power, you get 1 in 10 million.
I decided to put those numbers to the test. I went ahead and downloaded the 2d facial recognition set called “Aberdeen” that is found here.
It contains photos of 90 unique college students from Aberdeen sometime in the late 80’s or early 90’s. What I did was take one photo of a gentleman named Adrian (seen below) and match him up to each of the 58 other white males in the data set.
In planning this experiment I made 7 checkpoints on Adrian’s face where I would compare alignment by aligning the pupils and resizing the eyeballs to match in order to incorporate head size. If alignment was exact or very close, I marked that checkpoint with a “Yes” on a spreadsheet (which I will share upon request). The checkpoints were the eyebrows, the tip of the nose, the subnasale, the crista philtri (top of the lips), the labiale interious (bottom of the lip), the gnathion (chin) and the bottom ear lobe.
After comparing all 58 white males, here are the numbers I got.
None of the photos matched on all 7 checkpoints. One photo had 4 checkpoints, and four of them had 3. Here are the totals below:
4 checkpoints – 1/58 or 1.7%
3 checkpoints – 4/58 or 5.2%
2 checkpoints – 16/58 or 27.6%
1 checkpoint – 28/58 or 48.3%
0 checkpoints – 9/58 or 15.5%
What I learned is that certain alignment checkpoints were much more common than others, which were rare. When pupils are aligned, there is a higher than average chance that the eyebrows will align as well. Since humans have similar head sizes, alignment of the chin was fairly common. What was rare was alignment of the other features. See below:
Eyebrows – 36/58 or 62%
Chin – 21/58 or 36.2%
Bottom Lip – 5/58 or 8.62%
Subnasale – 4/58 or 6.89%
Top Lip – 3/58 or 5.17%
Nose Tip – 2/58 or 3.44%
Bottom of Ears – 2/58 or 3.45%
Multiplying the fractions you get 181,440/2,207,984,167,552 or
1 in 12,169,225 chance that all 7 facial features will align. Mark guessed pretty damn close.
Now of course I understand that 58 can be considered a small sample size, and it is. But I believe 58 random college students (who are of a similar age, race, and ethnic background) is a decent enough sample size to give you a ballpark estimate. Using a confidence interval calculator, we can be 95% sure that the average facial alignment checkpoint mean will fall between 1.08 to 1.54.
In the future I would like to do the same experiment but only between father and son or (non-identical) brothers to see what alignment percentages we should expect. In the meantime, let’s take a look at the number of famous people in the USA and how many we should expect to align on all 7 features.
This article on Deadspin tried to guess the number of famous people in the USA and they landed on 10,000 (with admittedly little research, but I think it is a decent ballpark estimate). Link
Now let’s include all the famous people who have died over the last 50 years. Let’s be liberal and double our number to 20,000. Hell, let’s make it 25,000.
According to the numbers above, there is a 1 in 487 chance that we will find just one matchup between two random celebrities that aligns on all 7 checkpoints. Hopefully this will help you look at our facial alignments in a different light. It will also force us to ask questions about why so many celebrities matchup with 5 or 6 different checkpoints so often, when it is so rare with random people. My suspicion is that 4 or 5+ checkpoint alignments are rare even amongst family members.