Registration for this round started in February 2017 and ended in December 2017.
After discarding bad and invalid entries (for example, those using disposable email addresses), this round had 7896 entries that were matched into 3948 pairs.
The matching survey for this round had 75 questions. The goal for matching was to put all persons into pairs so that the difference in personality between the members of each pair was minimized.
Difference in personality was measured by Squared Euclidean distance between the two peoples answers for all 75 questions. So, a higher calculated difference means more difference in personality. Zero distance would mean the two people had answered every questions the same. Matches were also restricted to being somewhat similar in age.
Not everyone could be given there most similar match, due to multiple people conflicting. Everyone ended up being guaranteed a match in the top 0.5% of people most similar to them. Some people gave such unusual patterns of answers that there were very few or no people similar to them. If matches had been randomly made, the distance in personality between each pair would have averaged 297. When matches were assigned, so that average distance was minimized, the best that could be done average 176. The graph below summarizes the quality of final matches that were made, showing the distribution of differences for the pairs that were actually mailed out compared to the distribution of differences if pairing had been done randomly.
Data from round one has now been deleted from the server to protect user privacy, the email alias system used for people to get in contact with each other will remain up for two more weeks. Summary statistics from the survey are archived here in the name of open-science. Individual level data can not be shared. A new version of the matching survey is up and the second matching round is now accepting registration and future rounds will require less participants so that waiting time is reduced.