RE: Exchange Transfer Transparency for 2016, 2017, and 2018
As I noted elsewhere when you have $40 million in blocktrades+alpha (and there may be other associated accounts) that are mostly opaque and you don't know what they are doing, this throws off the entire dataset. This isn't a $40 slop factor, it is a $40 million slop factor. That is enormous.
Which transfers (beyond blocktrades providing non-exchange services and personal transfers on the same account) do think are characterized incorrectly here?
I don't know because the blocktrades collections of accounts are opaque. There may well be exchange transactions not shown as such and non-exchange transfers shown as exchange. As you say the blocktrades account is included in the web site. Apart from the sheer magnitude of at least $40 million, the fact that it is used a lot means that getting it wrong can carry over and bias a lot of the data.
The total amount of value people take out (or put in) to the Steem blockchain should matter to us all.
There is no such thing as putting in or taking out of the blockchain (other than burn I suppose). All that can be done is to transfer between accounts. Exchange accounts have some properties that relate to buying and selling but sometimes not. You might send to an exchange and leave it there (or vice versa), or send to exchange from one account and from exchange to another. You haven't put in or taken out value in any of these steps, just moved it between accounts.
Of course at some level you just did a SQL query (with certain embedded assumptions, which are transparent) and here is the output. I'm not disputing that part of it at all. Just that when you are trying to accomplish something beyond the literal query itself, the mapping between the SQL query and meaningful conclusions is important (and up to the reader).
If the data is actually inaccurate because of a bug, I'll fix it.
I don't know if the there is a bug. As we discussed, the aggregates on the steemit accounts don't pass sanity checks relative to things like headcount and burn rate, which says to me that something is wrong. Maybe it is just a matter of the blocktrades accounts issue, or maybe something else, I don't know.
Another possible explanation would be additional exchange accounts (possibly lesser-known and/or not public) missing from the list, which means transactions to and from those accounts are also missing. Again, I don't know the exact flaw but it is clear from both perspectives (top down and bottom up) that there are major inaccuracies here (not inaccuracies in so far as the output of the SQL query, but in the numbers actually representing an accurate aggregate of what the title says they do.
I disagree. It may throw off comparisons between accounts used as exchanges for buying and selling, but I'm not trying to make conclusions about any such comparisons.
Yes, I understand this, but exchange accounts clearly do have special meaning and are even listed in some accounts like Vessel. I think that meaning is important enough to highlight and people do "deposit" or "withdraw" from exchanges all the time for all blockchains. Just because the blockchain itself didn't burn or create tokens doesn't change the meaning of the interaction as commonly understood.
IMO, if people hold money on exchanges, they don't own the cryptocurrency any more. I see this as RULE #1. If they move to an exchange and then back off an exchange to a different account, the data would accurately show that as one withdrawal and one deposit.
Well, if it's something specific leading to a problem, I'll happily make improvements.
Without knowing the exact flow, it's hard to make your input actionable. The data, as far as I know, does represent the sum of the daily net USD value of transactions to and from known exchange accounts. Any conclusions beyond that are up to others to figure out.
Okay, we disagree then. IMO when there is $40 million of slop it throws off most meaningful interpretations here. Yes we can say that such and such account sent or received $X worth of tokens to 'known' exchanges, but we can't really say how that compares with other accounts or the totals. If said account interacted with the blocktrades-related accounts (and/or other unidentified exchange accounts) in some manner we don't understand, it may be that the $X going to 'known' exchanges is very, very different from the amount going to actual (known and unknown) exchanges.
There is such a thing as error bounds, and the error bounds seem to be at least $40 million in magnitude. That is simply enormous relative to the flows going on here.
i would agree with that. Where we may disagree is how much additional interpretation is necessary to reach any sort of meaningful conclusions is required, or if it is even possible.
It may in fact not be actionable. The Steem blockchain has a degree of transparency but as we have seen from the results, being your best effort and I'm sure a sincere one, not passing basic sanity checks, it is not clear there is sufficient transparency to reach meaningful conclusions. Or perhaps if some additional mysteries can be untangled the gap can be narrowed.