Quote from: rpietila on March 29, 2014, 06:56:23 PM

Quote from: Carlton Banks on March 29, 2014, 04:58:01 PM

Quote from: rpietila on March 29, 2014, 03:00:25 PM

[...]
I think the evidence points out to 700k-1M users but would like to hear comments on the methods and not just "voting on the results" Wink

What evidence is there that the Mt Gox leak was complete? There are a few potential reasons for the hacker to release an incomplete dataset, plus they didn't necessarily get the whole dataset. How many accounts with all zero balances were in the data? Why were transaction history and full personal details not released? Clearly not all types of account data were leaked, which implies either incomplete access or incomplete disclosure.

- The addresses in the leak contained 0.9 million coins, which agrees well with the general understanding about how much there was at Mt.Gox
- Everybody can check their balance; I haven't heard reports of missing balances, I have heard that the balance has been correct
- Zero balances are omitted, as they are in my "# of bitcoinusers calculation"
- The distribution of coins in the balances (see my analysis upthread) satisfies some intricate statistical requirements that would have been difficult to fake

That's fairly convincing. Although you are relying on a secondary assumption, that Gox accurately reported their total for BTC obligations in full. They're the sole source of that figure, and they're proven to be unreliable in many ways (most intriguing is the apparent contradiction between how many coins they lost to fraudulent withdrawal claims and the recent analysis that has appeared in the press).

Still, it would be unlikely to remove a significant number of accounts without significantly altering the distribution according to your analysis of the blockchain as a whole. That's the strongest argument that the data is more or less complete, for sure.