Rollingmelons / Notebook / Randomness Analysis #data #science
#video (5)
#comedic (4)
#science (4)
#web-app (4)
#data (3)
#my-songs (3)
#tool (2)
#art (1)
#game (1)
#quiz (1)
#story (1)
#writing (1)
Randomness Analysis

An embarrassingly long time ago, I ran a randomness survey and posted it on reddit to get a nice, fun bundle of data. Here, just like I did with my color survey, I will share out the results and discuss them.

The point of my randomness survey was to see how badly humans suck at being random. It's a topic that's been studied and I'm not hoping to find anything new, I just thought it would be neat to have my own set of data. I asked my redditian participants to be as random as possible as they performed a few selections, but inevitably some patterns emerged.

The survey tasks went as follows, and I've also provided my expectations for each:

First, a random letter: I expected there to be some sort of pattern related to the QWERTY keyboard, probably with the central letters being selected more often.
Next, five random letters: Same as above, but with the additional expectation that people avoided picking the same letter twice (repetition avoidance).
Now, a digit: I expected numbers toward the middle to be picked more often, but not the EXACT middle.
Now, five digits: Like with the second prompt, I expected people to avoid repeating digits, even though a repeat is just as likely as any other number.
Click four times in the box: The most interesting task, in my opinion. I expected people to avoid the edges of the box because those areas are harder to click, but beyond that I wasn't sure what to expect.
Pick a direction: Sort of like clicking in a box, but more constrained. Trying to see if there is a preferred direction.
Pick a random basketball in this row: Another preferred-direction test, but with only one dimension. I expected it to basically be the same as the single digit test, since the digits are in a row on the keyboard.
Random word: I added this question a bonus because I knew I wouldn't be able to get enough data to really find anything. Only a couple people skipped it, even though it was optional.


One Random Letter

These letters seem easy to press with the right hand?

Five Random Letters

The middle row is popular again, and the bottom row of letters is very unpopular. Here's a breakdown:

Never end with the letter U, apparently. Here's a combo graph for your butt:

As far as repetition in the five-letter sets, 45 sets included the same letter at least twice (whereas 103 sets would be expected to repeat a letter in a truly random sample) and a letter appeared directly after itself 19 times (47 would be expected).

Random Digit

Actually, I'm just going to put this one on a graph with the next one.

Five Random Digits

The extremes were unpopular, and the letter 6 was really popular as a single-digit selection for some reason?

Here's the breakdown:

Regarding repetition in five-digit sets now...120 sets included the same digit at least twice (216 sets would be expected) and the same digit appeared consecutively 65 times (124 times expected).

Click Four Times in a Box

There was a tendency to click in the center of each of the four quadrants of the square, and there was a tendency toward the middle (which is to be expected; the edges were harder to click).

I simulated the same number of random clicks as in my sample with a script that legitimately clicked randomly (at least, as random as computers can be). It's pretty easy to tell which of the following data sets it is, huh?

What I find REALLY interesting is that the top left corner was the overwhelming choice for the first click.

Additionally, 86 sets of four clicks were "quadrantized", which is a word I made up that means exactly one click was in each of the four quadrants of the square. 30 quadrantized sets would be expected from a truly random selection.

And one final thing: The clicks in the survey were generally further apart than a random set would be. The survey clicks averaged 126.67 pixels between the two closest points, but I ran a random sample of 321 sets of four clicks ten thousand times and the average distance between the closest two points was 105.93 pixels. (I had to run a script because I couldn't figure out the math at all. What an interesting problem this is! For anyone who would like a math puzzle, you should help me figure out what the average minimum distance is between points on a plane when four points are selected.)

Pick a Direction

Meh, I can't offer much insight on this. Good luck determining what this picture means in your life.

Pick a Basketball

This is about what I expected. People didn't click the basketballs furthest from the center very much, but they also didn't click the basketballs in the exact center as much as they clicked the ones slightly off to the side. The middle of something doesn't seem as random to most people; when asked for a number between 1 and 10, 3 and 7 are the most common responses, not 5.

Random Word

This survey gave me the amazing opportunity to read 321 people's mostest randomest words. About two-thirds of them were nouns, and about one-fourth of them were either animals or food. (There seems to be a tendency to just spout out names of various animals and foods when trying to be random, perhaps with some colors thrown into the babble. And sometimes words are common random words because they just sound goofy, like "spoon".) Dirty words and abnormally long or obscure words also made up significant portions of the responses. "Penis" constituted five individual responses, naturally. Twenty-one words in all were repeated, including such fine words as "cat", "poop", "potato", "swag", "the", "random", and "optometrist" (???).

You can see the rest of the words and all the other raw data here.


People aren't random.

Thanks for reading.

Type the letters:
Hey, I can't read those letters