What Block Chain Analysis Tells Us About Bitcoin

In this piece, CoinDesk contributor Tim Swanson takes a look at the latest block chain data in an attempt to draw new conclusions about how bitcoins are being spent, why they're being held and what this may mean for the burgeoning market.

John Ratcliff is a principal engineer at NVIDIA who has a unique hobby, using 3-D tools to visualize block chain analytics. Over the past several months he has created visual aids for the community, in part to help show trends and highlight uncertainty.

For instance, below is a pie chart of UTXO (unspent transaction outputs), better known as bitcoins and their distribution by age. That is to say the amount of bitcoins based on their last use.

Here's his original chart from January:

The following chart is an updated version based on the bitcoin block chain as of 30th April.

As you can see, the oldest coins have still not budged or moved in the subsequent months for reasons described in a recent paper (pdf).

Investor impact

The most surprising area of growth is the three-to-six month section.

While there could be a variety of reasons for why this is now the largest segment (e.g., lost, stolen tokens, Goxcoins), the timing coincides with the November price bubble that peaked at $1100 per bitcoin - it has since fallen to roughly $450.

That is to say, these holders could and likely did buy during the peak and are simply underwater. If that is the case, rather than cashing out and realizing a loss, they are holding them and waiting (with the hope) that there will be a rally in prices once more.

Some readers may be thinking that two data points are not enough to draw a conclusion. For instance, what does the block chain distribution look like over the past year?

After exchanging some thoughts and theories with Ratcliff, he put together a very interesting animation of the daily changes in distribution based on age starting from the genesis block on 3rd January 2009 to 14th May (here is the raw data). Spoiler alert: not much happens the first year.

At the beginning of January 2014, the three-to-six months segment represented only 6.8% of all tokens.

Now it represents more than 3 million UTXOs in part because people probably do not want to unload because of the ongoing bear market.

However, the activity that is generally associated to commerce happens solely on the left side of the bar graph, in the period of a week or less, representing roughly 10% of all mined bitcoins. Yet as far as how much is related to day traders, gambling sites (e.g., Satoshi Dice, Prime Dice), remittances or other activity that cannot be determined by this method.

In Ratcliff’s view, "we know what and where the liquid bitcoins are. And what is interesting about the graph is that it shows the difference in liquidity at different time periods".

He is currently working on releasing a number of other ways to visualize this type of data beginning with the genesis block (i.e., showing every token on every address in 3-D). In addition, he has another shorter clip from 9th March 2013 through 12th May 2014.

Cautious conclusions

Yet, he says that you cannot definitively jump to conclusions from it.

For instance, if Bob moves 100,000 bitcoins, outside observers do not know what Bob was doing (moving a wallet like Bitstamp did in November, conducting real commerce, betting, etc.).

In Ratcliff’s view, his approach has one key distinction, if Bob has a bitcoin address and he spends one bitcoin (UTXO), observers will view that event as if Bob moved just one bitcoin – Ratcliff is marking age as age of last spend transaction.

This then proves that Bob still controls this address (an address which represents inputs and outputs) - this address is not dead which is a subtle distinction.

To him, what you can see is X% of bitcoins that are liquid (moving on a weekly basis), X% that are being "saved" and X% (or in this case, 30%) that may be gone forever. It is unclear if someone such as Satoshi Nakamoto (who is alleged to control approximately 1 million bitcoins) still controls the keys or how much market uncertainty this will create.

For additional analysis, I contacted Jonathan Levin, co-founder of Coinometrics for his insights on this tool and approach. He explains:

"The underwater explanation for the three-to-six month period is plausible but the problem with the method is that this measure is highly susceptible to changes in the exchanges and so inferring behaviour of individuals is difficult. If Bitstamp has not done another reshuffle since the one last year that would probably show up in the three-to-six month section.





I am interested in the one month and one week decreases as well. It seems that less bitcoins are being moved around the block chain now rather than in January. We have also seen volumes on the exchanges drop since December levels which may explain some of this drop."

Alternative approaches to analysis

This picture has also been making the rounds on social media and on CoinDesk:

The original source is a Technology Review infographic.

Yet, the data source mentions Sarah Meiklejohn, author of one of the best papers in this space, "A Fistful of Bitcoins" (pdf). I contacted her and she said that the data in question comes from Figure 3 in the same paper.

However, the likely explanation to the infographic is that mining pools and farms typically must sell off their bitcoins to cover their operating costs which makes it difficult to isolate actual commerce from mining activity.

What the graphic probably visualizes is the competitiveness and professionalization of the industry.

Virtually all miners have to spend their tokens within a month of mining them. This is not real economic activity for the ecosystem as nearly all of those funds are converted to a foreign exchange (fiat currency) and then paid to a utility or hardware company, this does not improve the protocol or its usability.

Another depiction of the network activity are these two charts:

The chart above is the on-chain transactional volume over the past year excluding the top 100 popular addresses (such as gambling sites).

The chart above visualizes bitcoin days destroyed (BDD) which basically measures old coins that move.

Basically, the older and larger the amount, the larger the leverage point (as statisticians call it). What these two charts show is that there is very little new movement in the chain.

One of the reasons this is worth mentioning is because at the beginning of 2014 there was an estimated 20,000 – 30,000 merchants that accepted bitcoins for payment. By the end of the first quarter, approximately 60,000 merchants accepted bitcoins. Yet there has been very little corresponding on-chain growth.

Instead, most of the growth is on the edges, in trust-me silos. The occasional spikes are outliers involving exchanges moving wallets from one wallet to another (such as Bitstamp did and Mt. Gox purportedly did).

Regarding BDD, Jonathan Levin said:

"It seems that aside from the big spikes which are ‘security measures, there are relatively few bitcoins that are being actively traded and moved around the block chain. Estimating this amount of supply comes from pie charts. I would give an estimate for this active supply of bitcoins by taking the six month figure.”

Block chain analytics and forensics is a nascent yet exciting field of study that will bring to light what kind of real activity is taking place on public ledgers. And trailblazers like John Ratcliff, Jonathan Levin and Sarah Meiklejohn are individuals to keep your eye on.

Visualization image via Shutterstock