Wanting perfect data is a fundamental problem, and not just for accountants. Following the question farther, if one wanted truly perfect data for example for ROAR or for AREA, one would not accept the word of the players or the tournament director, one would have to witness the event, at least the handshake at the end or get a video of the playing. That would improve the quality of the data, but not to perfection; the meaning of the handshake could be misunderstood (the winner looked glum and the loser happy) or the person in the video might not be who it was said he was. That brings up the deeper point: although it seems "free" data quality has a cost, and getting better quality generally costs something. The cost of perfect data is usually "a lot," and any data, not just ROAR or AREA, probably is probably not perfect. The question then is whether the data is good enough for the purpose it is used for.
Regarding the "duplicate accounts," some are not in fact duplicates, which is a non-trivial to sort out. I am willing to let them persist until someone notifies me because determining whether the accounts are truly duplicates rather than coincidences has a cost. If only people were required to have unique names ;-)
JR