NEW GAME |
RESUME GAME |
HOW TO PLAY |
ABOUT |
SETTINGS |
high score
NEW GAME |
RESUME GAME |
HOW TO PLAY |
ABOUT |
SETTINGS |
each round shows you a scatter plot: a cloud of one hundred points, each one a pair of two variables. your job is to guess how strongly those two variables move together, expressed as a number between zero and one. type your guess into the box and hit guess (or press enter). the closer you are to the true correlation, the more you are rewarded.
you start with three lives. guess within 0.05 of the true value and you earn a life back plus five coins; guess within 0.10 and you collect a coin; miss by more than 0.10 and you lose a life. string good guesses together and bonus coins start rolling in. the game ends when you run out of lives, and your best coin total becomes your high score.
correlation measures how closely two things rise and fall together. the number you are guessing is the pearson correlation coefficient, usually written as r. a value of one means a perfect straight-line relationship: knowing one variable tells you the other exactly. a value of zero means no linear relationship at all: the points are a shapeless cloud and one variable tells you nothing about the other. everything interesting happens in between.
in the real world correlation can also be negative, where one variable goes up as the other goes down. to keep things simple, this game only uses positive correlations from zero to one. estimating a negative correlation is exactly the same skill, just mirrored, so what you practise here carries straight over.
the single most useful trick is to picture a straight line running through the middle of the cloud, then ask how tightly the points hug it. points packed close to that line mean a high correlation. points scattered in a fat, fuzzy band mean a low one. a perfectly round blob with no direction sits near zero; a thin diagonal streak sits near one.
a common trap is to judge the slope of the cloud instead of its tightness. correlation has nothing to do with how steep the relationship is. a gently rising line and a steeply rising line can share the exact same correlation, because what matters is the spread of the points around the line, not the angle of the line itself. train your eye on the thickness of the band, not its direction.
• calibrate against the examples above before you start, and keep a mental picture of what 0.3, 0.6 and 0.9 actually look like.
• the middle range from about 0.4 to 0.7 is the hardest to call, so spend your practice there rather than on the easy extremes.
• don't let one or two stray outliers pull your guess around; judge the bulk of the cloud.
• most people systematically underestimate strong correlations, so if a cloud looks tight, nudge your guess a little higher than feels natural.
• work quickly and trust your first read. staring longer rarely improves the guess and the timer is running.
guess the correlation is a game with a purpose. how people perceive correlation in scatter plots is a genuine question in perception and data-visualisation research: our eyes are not calibrated instruments, and we read the same chart in surprisingly different ways. playing the game is a fun way to sharpen a skill that matters any time you look at data, whether in a news chart, a research paper or a spreadsheet at work.
where do the scatter plots come from?
every round is generated fresh in your browser with a randomly chosen correlation, so no two games are the same and there is no pattern to memorise.
why are there no negative correlations?
purely to keep the game approachable. a negative correlation is just the mirror image of a positive one, so the skill is identical.
is any of my data collected?
no. the game runs entirely on your device. your username, high score and progress are stored only in your browser and are never sent anywhere.
what counts as a good guess?
landing within 0.05 of the true value is excellent and earns you a life and five coins. within 0.10 is still solid and earns a coin.
i lost all my lives, how do i start over?
just hit play again on the game over screen, or head back to the main menu and start a new game.
guess the correlation is a simple browser game about estimating how strongly two variables in a scatter plot are related. it runs entirely on your device, and your username, high score and progress are stored only in your browser. new to it? head to the section for a full walkthrough and tips.
i developed this game while i was a phd student studying bioinformatics at the university of cambridge and the european bioinformatics institute. this game is a side project to feed one of my many day-to-day curiosities.
i'm always grateful for suggestions and happy to answer questions about the game. so tweet me at @omarwagih or email me.
warning: this will erase your username, highscore and game progress stored in this browser