Unless they had several thousand couples for each one of the 144 cells, I’m very surprised there weren’t bigger fluctuations due to chance alone. (And that single “59” shows that they didn’t round all numbers to the nearest ten.)
Sorry, I should have linked the article earlier instead of just the chart.
On sample size: Keep in mind that it isn’t couples that are being looked at here, just comparisons between users’ self-reports. Specifically, each question has two answers: The user’s self-report, and what they would want a potential date to answer. The compatibility percentage is based on matching from A’s wants to B’s reports and vice-versa.
For the article, data was collected from a randomly selected pool of 500,000 straight users. The gender balance among straight users is about 60% men, 40% women, so that’s about 25,000 men in each row and 17,000 women in each column. So each cell has about 400 million comparisons.
Indeed they did—about 868 million couples per cell by my reckoning, or about half that if they’re only pairing based on preferred gender:
Here are the grouped match percentages for a random pool of 500,000 users.
Astrological sign has no effect whatsoever on how compatible two people are.
[...]
We’re showing you this table, as dull as it is, because the uniformity neatly
illustrates how beefy our data set is. There are 144 pools considered above,
and they all match the mean plus or minus 0.5%.
Unless they had several thousand couples for each one of the 144 cells, I’m very surprised there weren’t bigger fluctuations due to chance alone. (And that single “59” shows that they didn’t round all numbers to the nearest ten.)
Sorry, I should have linked the article earlier instead of just the chart.
On sample size: Keep in mind that it isn’t couples that are being looked at here, just comparisons between users’ self-reports. Specifically, each question has two answers: The user’s self-report, and what they would want a potential date to answer. The compatibility percentage is based on matching from A’s wants to B’s reports and vice-versa.
For the article, data was collected from a randomly selected pool of 500,000 straight users. The gender balance among straight users is about 60% men, 40% women, so that’s about 25,000 men in each row and 17,000 women in each column. So each cell has about 400 million comparisons.
Indeed they did—about 868 million couples per cell by my reckoning, or about half that if they’re only pairing based on preferred gender: