DASH stats 1: London’s “Buyout” problem

Lemonade stand

Chris has already written about the DASH audio I’ve put together. The best audio deals with topics that work as audio. Clips of what it was like to be there, experiences of solving puzzles, that works in your ears. Hardcore statistical nerdery, that needs to come out through your eyes.

In a later post, I’ll be looking at the difference between the New and Experienced tracks, from analysis of the results. Here, I’m considering whether the UK players in DASH 6 were disadvantaged by one of the puzzles.

London had a problem. Players in Britain mostly spoke in British English. It’s a dialect similar, but not identical, to American English. Spellings alter, words have other meanings, and there are major differences in commercial culture.

These differences came to the fore in the “Buyout” puzzle, which attempted to clue to a commercial company. It’s an American brand, unfamiliar in Britain. The organisers had attempted to reduce this problem, by substituting a similar but different puzzle, but this hadn’t quite worked out.

London players experienced a different puzzle, each clue resolved to a verb-noun combination. Other players got a verb-noun combination that makes a commercial product. Were London players at a disadvantage by not getting the branded nature of the puzzle?

A Student’s t-test is appropriate to compare the performances between London (where 13 teams took New and 8 Experienced) and All Other Locations (84 New, 300 Experienced). The t-test compares the difference of each value in a set from the set’s mean, and works out the probability that the two data sets have the same mean – the “null hypothesis” in this test. It’s especially useful for sets of different sizes.

Technical points: I’m only considering those teams that recorded a solve time for “Buyout”. Also, note how London had more New than Experienced teams: I can conjecture that, if anything, London teams may have under-estimated their abilities, and could prove better than All Other Locations.

For the “Buyout” puzzle, analysis suggests that I can reject the null hypothesis, and there probably *was* a difference between London teams and all others. On the Easy track, there’s a 3% chance that the observed scores come from the same population. These probabilities are even lower for the Experienced track, where the small sample set might not be representative.

If we boost the scores of London teams by 2 points, we can accept the null hypothesis on all measures. So, yes, the teams of London appear to have been disadvantaged by the cultural gap, but only by a minute.

As a control, I’ve repeated this analysis for the other puzzles. We can easily accept the null hypothesis in all cases, and assume that there was no trans-atlantic difference.

3 Comments

  1. The post is interesting but a little too statistically technical for me to understand fully. As part of an ‘experienced’ UK team, I can unequivocally state that the finish to Buyout cost us considerably more than a minute (more like 10), involving as it did the identification of a US brand name unknown in the UK. Obviously there’s a chance that the preliminary part of the puzzle, for which we were given different clues to the US, for different answers, may have swung things either way – the UK probably had easier clues, but to a less coherent answer set (and one that didn’t give much of a hint about the meta-answer). We had plenty of other struggles during the hunt, but this was a very frustrating one, which we felt cost us quite a few places on the leaderboard.

    Reply

Leave a Comment.