Case 1 · skewed data · median v mean
Half our games get 4% of the answers
When we decided (12 November 2026)
1Ask
We had time for one big job before Christmas: build new games, or help people find the 116 we've already got. Which one would get more students answering questions?
2Collect
Every game counts the answers given in it. From 1 September to 12 November: 25,438 answers across 116 games and rooms.
The six with the most answers
Show the figures as a table
| Game or room | Answers |
|---|---|
| Six Sevens, Bruv | 4,181 |
| Like Terms Collector | 2,475 |
| Prime or Composite | 1,930 |
| Factor Race | 1,882 |
| The Kiln Disaster (room) | 775 |
| Negative Number Line | 729 |
… and 110 more. The bottom 58 have 987 answers between them.
3Picture it
Line all 116 up from most played to least, cut the queue into ten equal groups, and see what share of the answers each group gets.
Share of all answers, by tenth of the games
Show the figures as a table
| Group | Share of answers |
|---|---|
| Top tenth | 59.9% |
| 2nd | 18.9% |
| 3rd | 9.2% |
| 4th | 5.1% |
| 5th | 3.1% |
| 6th | 1.7% |
| 7th | 1.1% |
| 8th | 0.6% |
| 9th | 0.3% |
| Bottom | 0.1% |
The top tenth of games (11 of them) get 59.9% of all answers. The bottom half get 3.9%.
- Mean answers per game
- 219
- Median
- 52
- Interquartile range
- 13 to 192
Which average tells the truth about a "typical" game? A handful of huge games drag the mean up to 219, but only 28 of the 116 games actually reach it. The median, 52, is what most games really look like. When data is this lopsided, the mean tells you about the giants, not the typical game.
Today
Has the bottom half moved since we decided?
4Decide
Both, but finding came first. Building another game would add one more to the bottom half. So in November we made the games we already had easier to find: every escape room got its own web address and a title that says what maths it covers, the site map was fixed so search engines can see every game, and every game's start screen now has a way back to the rest. We're still building, but new games now have to earn their place. We'll run these numbers again in January: if the bottom half is still getting 4%, we were wrong about why.
What this data can't tell us
- Answers aren't students. One keen class can give a game 500 answers in an afternoon.
- Not every game had the same time. A game launched in October had weeks less than one live in September.
- Some answers are ours. We test games by playing them, and those answers are in the count.
- Popular isn't the same as good. A game might be quiet because nobody needs that topic yet this term.