Ten options, and nothing close to ten percent each. Asked for a number in a small range, people behave consistently enough to bet on.
Think of a number from 1 to 10.
The first one that comes to mind. Press it.
Seven, and this one has real measurement behind it. Kubovy and Psotka asked 558 people on the Yale campus for the first digit that came to mind, and 28.4% said seven — The predominance of seven and the apparent spontaneity of numerical choices, 1976.
A British replication decades later found the same thing a little stronger: YouGov asked for a number from one to ten and 33% chose seven, with people avoiding one, two, nine and ten. An even spread would put every number at 10%.
The two questions are not identical — the Yale study asked for a single digit, this page asks for one to ten — which is part of why a direct count is worth having.
Out of 0 answers so far. Free to quote, with the sample size beside it so a reader can judge the weight.
Once this passes 1000 answers, the measurement replaces the claim above as the headline of this page, and the claim becomes the thing being checked rather than the thing being reported. Saying that in advance is the point: a threshold decided afterwards is not a threshold.
1 and 10 are available and hardly anybody uses them, which is what YouGov found. They are the boundaries, and a boundary does not feel like a choice — picking 1 reads as refusing to pick, and picking 10 is the same gesture from the other end.
That removes a fifth of the range before anybody has consciously considered it, and it happens fast enough that most people would deny doing it.
The next filter is stranger: people asked for an arbitrary number reach for odd ones. Even numbers feel structured, halvable, tidy. An odd number feels as though it came from nowhere, which is what somebody trying to be unpredictable is aiming for.
Between the two filters, 3 and 7 are left standing in the middle of the range wearing the best disguise, and 7 has one more advantage: it is the only number from 1 to 10 that cannot be reached by doubling or halving another number in the range. It is the most awkward number in a very tidy set, and awkward reads as random.
The same bias appears wherever people are asked to produce randomness rather than measure it: in survey answers, in invented data, in the digits chosen for a PIN. Humans are poor random number generators in a specific and repeatable direction, and a performer exploiting it is doing applied psychology with a deck in hand.
The shuffling page is the other half of the same idea: how much shuffling it takes before a deck stops carrying the order a person gave it.
Seven. YouGov asked Britons to write down a number from one to ten and 33% chose it, against the 10% an even spread would give.
People also avoid the numbers at either end — one, two, nine and ten were all unpopular in the same survey.
Because a boundary does not feel like a choice. Picking the first or last option reads as declining to choose rather than choosing.
That alone removes a fifth of the range before anybody has thought about it.
Reliably so, and in a specific direction rather than at random. People avoid the ends, prefer odd numbers, avoid repeating themselves, and avoid whatever seems too obvious.
That is what makes a guess worth making, and it appears well beyond magic — in survey answers, invented data, and chosen PINs.
This is one of eleven questions in the full study, which asks for a card, a number, a vegetable, a fruit, an animal and more, and shows the running counts for all of them. The rest of the site is self-working card tricks you can play in the browser.