Quote from: k on August 31, 2011, 10:43:05 PM

Hi fergalr,

thanks for your paper and your responses here. Welcome to the forum.

Thanks!

Quote from: k on August 31, 2011, 10:43:05 PM

Have you given any thought to other applications these tools and this type of network analysis could be used for - for example could it be used somehow to estimate the real size of the bitcoin economy, i.e. differentiate the purchase of goods and services from just shifting bitcoins to different addresses owned by the same person?

These are really interesting questions.

We thought briefly about these issues, although our main focus was on anonymity, and once it became clear to us that users were less anonymous than we expected, on making our analysis public.

So, there is a large connected component (http://en.wikipedia.org/wiki/Connected_component_%28graph_theory%29) of Bitcoin addresses - i.e. a set of addresses that are all connected via other addresses, where 2 addresses are connected if they have exchanged bitcoins.

Bitcoins in accounts outside that large connected component could probably be discarded from any attempt to estimate the velocity of money (http://en.wikipedia.org/wiki/Velocity_of_money).
There are definitely examples of this - where someone consolidates a large bunch of mined bitcoins, but then doesn't do anything else with them. You'd really want these consolidation events to not show up in an analysis of the velocity of money, and hence the size of the economy - so you could probably do some work on this - a crude first pass being to just look at Bitcoins within the large connected component.

So, that's one type of analysis you could do.
The other thing is that we typically 'link' all the addresses that the block chain reveals are controlled by a single user (because their private key parts are used in a single transaction). This collapses many addresses together - from something like 1.2M unique addresses down to to .8M 'collapsed nodes'.
Now, there is an issue here, that occasionally some key management services, like myBitcoin, where the private keys are controlled centrally, show up as a single node, in this graph - and obviously can have a lot of different users.
So you basically lose transactions that occur within a single 'virtual' service, that are backed onto the Bitcoin network.

But for the vast majority of nodes, you get a network that much more resembles the user->user transaction network.
So, its not perfect - there will definitely be accounts that are controlled by a single user, that still show up as multiple accounts, and there'll be some accounts, where users use a service that shares their private keys, that now look like a single account - but it's certainly a different, and interesting, view of the network, on which to do economic analysis.

From looking at it, I'd say its a better view, and that things like velocity of money calculated on this network, would be more accurate.

The other thing I should say is that there were some forum users here that built a 'bitcoin days destroyed' metric that is also interesting.
Some combination of the two ideas might be beneficial, in terms of analysing the economy.

You could probably build further heuristic methods - to look at the volume of Bitcoins that flow a certain distance, through the resolved nodes, and use this as a better barometer of the current market activity.

Quote from: k on August 31, 2011, 10:43:05 PM

Or identify the different exchanges and see the flow of bitcoins to them and thus if you see a larger than normal flow of bitcoins to a particular exchange it might indicate that a sell-off is likely and predict a price drop before it happens?

That sort of stuff is very interesting - things like that are probably possible.
Like, if nothing else, you can look at large volume movements that happen outside exchanges - its probably possible to correlate such movements with increased probability of near future volatility, if nothing else.

Quote from: k on August 31, 2011, 10:43:05 PM

Maybe nothing like this is possible, just thinking out load and wondering about possible other uses for this type of analysis.

thanks
k

From my point of view, while we've thought a little in that direction, it is a whole other research topic!

But it definitely sounds like a very interesting one - especially if Bitcoin gains momentum, and people start doing serious trading on it - I like the direction you are thinking in.