The Myers-Briggs places weight on questions to give one of the 16 personality types made of 8 traits. For example, one answer might give you 80% of INTJ and 20% INTP and by factoring the weights across all the questions will tell you if you were weighted more or less towards INTJ or INTP, even if you were more INTJ for that singular question.
In 8x8 the questions are not weighted in that way, which is an important difference. While there are 16 distinct personality traits in MBTI, there are only 8 traits. Questions can weigh towards those 8 traits in various ways. From my experience MBTI tests range from the 5 "strongly, disagree, neutral, agree, strongly" answer types and the questions are targeted to 2 (sometimes 4, but typically 2) traits.
Meaning if neutral is 50% : 50% (Not leaning towards introvert or extrovert) your answer to the question is effectively meaningless. However if you "agree" it might be 75% : 25% and if you "strongly agree" it could be 100% : 0%. You take the aggravated ratings of all the "I or E" type questions and determine if the user is introverted or extroverted based on which holds more weight across 50 questions.
8x8 does not weight the questions to determine personality traits. The questions must match exactly to determine a match. In comparison, Myers-Briggs can have two people be INTJ who answered questions very differently but their aggregated weights of each question added up to them both being an INTJ.
Again: this should be treated as a case study. Since questions aren't weighted, this is different than most personality tests. This isn't a personality test - but finding someone who thinks exactly as you do in regards to these 8 questions.
The largest flaw is if these are the wrong 8 questions to ask and if matches lean towards "good" or not. Which can be determined after enough matches. Making this a neat case study. :)
E:
TL;DR If 12,000 people took a MBTI I guarantee there would be many matches already. This isn't a personality test - this is a line of thinking test. Whether the questions will result in good matches or not will be left to when there are any matches to begin with.
And I highly doubt they have no matches in 10k samples, especially given the homogeneity of their audience. And if so then they are just measuring traits of personality that, while diversified, nobody cares about. Check out the birthday paradox. Even if there are 16M possibilities, the chance that two random people in 10,000 matched is much higher than you think.
Until the results speak for themselves (ie. accuracy of matches) you have no way of knowing if this is a complete failure or successful venture. Although you are free to speculate that it will be meaningless due to the methodology, you have no evidence to back that up yet.
Many dating sites work on a similar principle, although they all use weighted questions. Weighted questions for personality increase possible matches by narrowing down the preciseness necessary. The goal here is to decrease possible matches by requiring exact matches.
Dating sites want to match you with someone quickly who is "close enough" to be compatible based on a personality test and their user data of what answers pair well with other peoples' answers.
8x8 doesn't care how long it takes to find a match, they want an "exact match" to be compatible based on how you answer arbitrary questions.
8 questions was chosen because to have answers be "100% exact" are already rare. If they chose 50 questions - the chances of an exact match become increasingly rare as to possibly never exist in a persons' life time.
E:
The birthday paradox is 1/365, and I'm well aware of the statistics of someone out of a group of 20 random people sharing a birthday with another person of that same group.
It is quite possible that they have not tried to do the match online. Given their demonstrated competence, I wouldn't be surprise if it comes as a shock to them that they can simply put results in a flat file, sort, and then scan them for matches very quickly...
There were 11,300~ tests when the page last loaded for me. They are caching the results of the tests and refreshing non-live time in an attempt to put less stress on the server.
There were 0 matches at that time. While I'm wary of site-counters - I'll give them the benefit of the doubt.
E: The formula is
P(N) = 1 - prod[ 1 - (i-1)/(88 ) , i = 1,2,3,...N]
At 15,000~ there should be a match. So while pretty high at 12,000~ it's still in the realm of possibility that there is not a match, just very unlikely.
There's also the chance they haven't had the ability to check for matches because their servers been down so often so they haven't been calculating for them. So there is a chance that there is a match in the data but it hasn't been calculated and so wasn't showing when I checked at 11,300~
(1 - reduce(operator.mul, map(lambda a: (8 ** 8.0 - a) / 8 ** 8, range(12000)))) * 100.0You mean, for 2 people? It's higher than that http://en.wikipedia.org/wiki/Birthday_problem
I really don't believe their number of matches. Why... We're on the Internet, it's full of trolls.
I'm sure any number of people will have taken the test at two different locations and answered the questions the same, just to get a match.
Check out http://en.wikipedia.org/wiki/Entropy_%28information_theory%2...
Information gain is maximized when the distribution is even.
Imagine a very extreme biased form of this where everyone answers the same way to a question, and that question is repeated 8 times with different wordings. You will only get 1 distinct personality type.
For a real-world example, take the myers briggs (MBTI) personality test. It only does 16 distinct personality types and takes about 50q to get right. Only way 8^8 beats that, without any data to predict the distributions and correlations out of the gate, or completely revolutionarily novel psychological theory, is with precogs.