How Many Sweets Are in the Jar?

weirdly accurate.

How Many Sweets Are in the Jar?

How do you guess how many candies are in a jar?

This candy jar guessing game puts the classic fairground puzzle on a turntable: five jars, each harder to read, and a count you guess before the sweets are tipped out. The usual advice is to work it out from the jar’s volume, the share that sweets fill and the size of one sweet. After every jar you see the true count, and how close the 812 people who guessed real jars in a 2018 study would have got.

This game is a climb of five levels: a small straight jar, a tall jar of flat sweets, a wide shallow bowl, a round fishbowl of big and small sweets, and a big jar of tiny sweets, from about 40 to 3,400 sweets. The sweets are dropped in one at a time, so the count is exact. You drag to turn the jar, tap your guess on a keypad, and the jar is tipped out and counted as the sweets land. A guess within 25% of the true count clears the level; a miss gets one more jar of the same level, and a second miss ends the run. After every jar you see how many people in Kao and colleagues’ 2018 study would have got that close, worked out from their fit for that jar’s count.

It is not a test of intelligence or eyesight; it is a game about estimating. A jar on a screen is not quite the same job as a real jar on a table, and unlike the study’s guessers you see the answer after every jar, so later jars let you learn. The comparison with Kao’s crowd is a model worked out for each jar, not a ranking of other players; the site stores no guesses.

What the game measures

One skill, five levels: how close your eye gets to the true count as the jars get bigger and harder to read.

The task

  • Five 3D glass jars of sweets on a porcelain turntable, from about 40 to 3,400 sweets: a straight jar, a tall jar, a wide shallow bowl, a round fishbowl, a big apothecary jar
  • Sweets get smaller and more mixed level by level: big round ones, flat ones, two sizes together, tiny ones of every size
  • Drag to turn the jar and tilt it a little, tap your guess on a keypad (up to 99,999), then lock it in
  • Within 25% of the true count clears a level; two tries per level, and the run ends on the second miss
  • Everyone who plays on the same UTC day gets the same jars, so friends can compare fairly

Your result

  • How many of the five levels you cleared, and every jar’s true count, your guess and how far off, as a percentage over or under
  • For each jar, how many people in Kao et al.’s 2018 study would have got within 25%, and on average how many you beat
  • Whether your guesses drifted one way as the jars grew, next to the crowd’s typical guess at three of the study’s jar sizes
  • How full of sweets your last jar really was, a shareable card with a scan code and a challenge link

Before you start

  • About two minutes, five levels
  • Mouse, keyboard or touchscreen; works on a phone
  • Sound is optional: short effects only, with a mute button
  • Needs a browser with WebGL (3D) switched on

Candy jar guessing FAQ

Is this test free? Do I need to sign up to see my result?

Yes, it’s free, and there’s nothing to sign up for: no account, no email, no payment. Your result appears on this page the moment you finish, and you can share it or play again straight away.

What is the best way to guess how many candies are in a jar?

Estimate the jar’s inside volume, multiply by the share that sweets fill, and divide by the volume of one sweet. Donev and colleagues (2004) found that randomly jammed balls fill about 64% of a space and M&M’s-shaped candies about 68%. Then average several independent guesses, your own or other people’s, because their errors tend to cancel.

How is the jar guessing game scored?

The jar guessing game is scored by how many of its five levels you clear. Each guess is your error as a percentage of the true count, and a guess within 25% either way clears the level. You get two jars per level; missing both ends the run. Every jar is also set against the crowd in Kao et al.’s 2018 study.

How close do most people get when guessing a jar?

Every jar answers that as you play: once the sweets are counted, the page shows how many people in Kao et al.’s 2018 study would have got within 25% of a jar that size. Those 812 people in two US towns each guessed one real jar of 54 to 27,852 objects, with a prize for the closest, and their guesses spread roughly lognormally. The result sheet averages the comparison over your jars.

Do people guess big jars differently from small ones?

The result sheet shows that after your last jar, because knowing first would change how you guess. Kao and colleagues’ 2018 crowd guessed jars from 54 to 27,852 objects, and their fit shows how the typical guess moves with the size of the jar. The sheet puts your own guesses beside the crowd’s typical guess at three of the study’s jar sizes and tells you whether you drifted one way.

Why are the jars the same for everyone today?

Each jar is built from the UTC day number, the level and the try, so everyone who plays on the same day meets the same jars and friends can compare fairly. A challenge link carries your result inside the link itself. The site stores no guesses.

Is the jar guessing game an intelligence or eyesight test?

No. The jar guessing game is about estimating, not intelligence or vision. One jar is a noisy measure, a screen is not a real jar on a table, and you see every answer as you go, which the study’s guessers did not, so the comparison with their crowd flatters you more the further you climb.

Where the numbers come from

Every comparison comes from the lognormal fit Kao et al. (2018) made to 812 people guessing five real jars (54 to 27,852 objects), worked out for each jar’s true count; the 25% line that clears a level is the game’s, not the study’s. The site stores no guesses and does not publish players’ data.

How it is scored

Error

Your guess minus the true count, divided by the true count, shown as a percentage over or under (or as a multiple when you guess more than twice the count). Within 25% either way clears the level; within 10% is called spot on.

The climb

Five levels in a fixed order of difficulty. Each level gives up to two jars: clear one and you move up; miss both and the run ends. Your score is the number of levels cleared, 0 to 5.

Compared with a published crowd

Kao et al. (2018) found jar guesses close to lognormal at every jar size. Their fit is worked out for each of your jars’ true count: how many of their crowd would have got within 25%, and how many you beat by ratio, so twice too many and half as many count as equally far. The result shows the mean over your jars, 1–99%.