Here’s a neat game to test your knowledge of statistical trends.

Created by the graphic designer Olivier Ballou. And here is a link to the website.


The final scores is always TOO big. If you just draw a line, it's still around 60-70% score.

The axes are relatively fixed. No negative values, no full range of options. Probably constrains the variability to an extent.

Is this the one where you can cheat by pretty much always guessing the counterintuitive answer?

Nah. It seems to mostly align with what you'd expect, even if you only know only a little bit about the topic.

I scored 115%; I know a lot about statistics.

Very impressive.

It looks like everyone else got 87%, but I, shamefully, only got 84%. Mostly because I didn't realize the foreign born population in the early 20th century was so high.

85% here.

Looking at the results above, you probably could have put in a straight horizontal line and gotten 70%

I checked, and it turns out you can actually get 75% that way.

Keeping the scale of the Y-axis 0-100% at every question would make it more fun, it just gives too much clues to topics unknown.

91%. Lowest score was on NYC murders -- I underestimated the height of the post 1970 spike.

If I am guessing correctly, then the score calculated is nothing but R^2 value.

It was interesting to see the fact they decided to highlight (not sure if it was the same for everyone). For example, I would have been more interested in what events occurred around the time when the murders in NYC started exploding or dropped off, not that crack made its way to NYC in the 80s while the prior trend continued.

So he's copied the NYT "you draw it" series.

86%. Turns out I don't know much about first marriage ages.

But seriously; broken scoring statistics. The entropy in the player line should be related to the entropy in the actual line; flat horizontal responses should be penalised unless the real effect is also flat.

89%. The NYC murder graph really surprised me.

Americans give an email address when asked for one? Weird!

Or did everyone use a burner address?

N=1 sample of Americans here: I use a address created specifically for online identity. It also adds to my affectation of being a Russian troll.

Comments for this post are closed