Exploring the Tangle
One of the most important features of a permissionless DLT is that every transaction is transparent to everyone. One might expect that this would make it easy to see who is doing what on the Tangle and how the network usage is developing. Nevertheless little is known about how iota’s Tangle is actually used. We know the number of transactions per second, how many of them are confirmed and how many have (nonzero) value. Check iotaprices.com for an up-to-date overview of those basic stats. Thanks to thetangle.org we can also view the inputs and outputs of any transaction or address. But that’s about it. Since I would like to know more about the real usage of the tangle, I delved into the data. What I am particularly curious about is what we can say about the users of the network. Can we distinguish between hodlers, speculators, developers and other economic actors? Is it possible to identify the flows to and from exchanges? Would it even be possible to say something about the number of users or the number of active services/applications?
Exchanges and value transactions
For the first analysis in what might become a series I started by exploring (chains of) value transactions. And because they always start (and usually end) at an exchange, I tried to specifically identify the activities of the exchanges. My assumption is as follows: a large exchange makes a bundle of transactions several times a day (probably even several times an hour). These bundles often contain a large input, an output in the same order of magnitude, and in addition one or more smaller transactions (collecting the deposits and disbursing the withdrawals). The largest output of such a bundle is in turn the largest input of the next bundle. With these principles it seems indeed possible to identify a number of series, at least as far as the withdrawals are concerned.
Binance
We can see, for example, a series of transactions starting on August 4 with 3.9 Ti, followed by the transfer of iotas with a high frequency (over 50 bundles per day) until the balance decreases to below 500 Gi after five days. During this period, 267 bundles of value were distributed to a total of 1017 addresses. This represents an outflow of 680 Gi per day.
As soon as the outflow is depleted, a new address is filled with exactly 5 Ti, after which the entire story repeats itself. Interestingly, the filling is done by one transaction that swipes 120 addresses (probably these are the deposits ) together and adds 500 Gi from an earlier withdrawal series. These addresses have sometimes been inactive for months, but sometimes only for a few days.
More about the inputs for the withdrawal series will follow later.
A test transaction shows that this series is the work of Binance. During the period covered by the study, there are more such sequences (probably other exchanges), but none of them take place as frequently and with such large values as Binance. The other major exchange, Bitfinex, uses several smaller hot wallets. More about that later.
So now we are able to calculate the Binance-withdrawal rate. In general there are between 100 and 200 withdrawals per day (unique addresses), with a total value of 1 to 1.5 Ti. Next steps will be to quantify the deposit-rate, for Binance and other exchanges. Deposits are more difficult to investigate because an address can only be linked to an exchange when that exchange merges the deposit with other deposits in its (hot or cold) wallet. But if and when we have identified deposits and withdrawals, we can explore what happens between a withdrawal and the subsequent deposit: are the iotas only sitting at an address (for a shorter or longer time?), or are there actually economic transactions?
If you have any suggestions for further research, please let me know!
