Birthday paradox, 8^8, etc.

[Dave Britten]

(05-19-2015 08:52 AM)Maximilian Hohmann Wrote:  Hello!

(05-18-2015 02:39 PM)Dave Britten Wrote:  I kind of figured the data would be far from a random distribution.

That would be my guess as well. Unfortunately I am not allowed to read the eight questions ("Sorry 8^8 is not yet available in your country at this time. Hopefully a translated version will be available soon." - what nonsense! I don't need a translated version...). I suspect that the eight possible answers will not be randomly distributed, but weighted by to the cultural context - only people from the US seem to be allowed to take the test. So for each question there will be two or maybe three answers getting the majority of clicks. This effectively reduces the 8x8x8x8x8x8x8x8 combinations to something like 2x3x2x3x2x3x2x3.

Regards
Max

That's kind of weird. You aren't missing too much as far as the questions, though. It's just basic personality/values kind of stuff.

I hope they release the raw data at some point for recreational statistical analysis. Just the 8 selected answers, and the user's country (obtained via IP address) would be all that's necessary.

It's also worth noting that we don't know exactly how they're counting "matches". When I ran the simulation, I counted any insertion collision as one match (10 with the same answers = 9 "matches"). But if they count two in one bucket as one match, three as three, four as six, etc. (i.e. polygon sides + diagonals, n+(n*(n-3))/2), then their numbers would be a lot higher.