Accepted Manuscript

Digital blockchain networks appear to be following metcalfe's law

Ken Alabi

PII: S1567-4223(17)30048-0
DOI: http://dx.doi.org/10.1016/j.elerap.2017.06.003
Reference: ELERAP 714

To appear in: Electronic Commerce Research and Applications

Received Date: 13 June 2017

Revised Date: 18 June 2017
Accepted Date: 18 June 2017

Please cite this article as: K. Alabi, Digital blockchain networks appear to be following metcalfe's law, Electronic
Commerce Research and Applications (2017), doi: http://dx.doi.org/10.1016/j.elerap.2017.06.003

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
 DIGITAL BLOCKCHAIN NETWORKS
APPEAR TO BE FOLLOWING METCALFE'S LAW

Ken Alabi
Stony Brook University
alabi.ken@gmail.com

Last revised: June 18, 2017

_____________________________________________________________________________________

ABSTRACT
An analysis of some of the recent blockchain networks is presented to determine if they satisfy Metcalfe’s
Law for networks, as has been shown online social media networks. The value of a payment network was
modeled based on the price of the digital currency in use on the network, and the number of users by the
number of unique addresses each day that engage in transactions on the network. The Bitcoin, Ethereum,
and Dash networks were analyzed. The analysis shows that the networks were fairly well modeled by
Metcalfe's Law, which identifies the value of a network as proportional to the number of its nodes, or the
number of its end users. A new network model is also presented that shows the value to be proportional to
the exponential of the root of the number of users participating in the network, and shows good agreement
as well. Conditions for determining critical mass based on the new model are also presented. Finally, the
potential for identifying value bubbles that can be spotted as deviations in value from the model is dis-
cussed and illustrated using the data from one of the networks. Those value bubbles show up where re-
peated extremely high value increases are not accompanied by any commensurate increase in the number
of participating users, or any other development that could give rise to the higher level of value.

Keywords: Blockchain, cryptocurrency, digital networks, distributed ledger systems, distributed transac-
tion processing, fintech, Metcalfe’s Law, network effects, online payments, payment networks.
_____________________________________________________________________________________

Brief bio of the author. Dr. Ken Alabi has a Ph.D. in Engineering from Stony Brook University and a
Masters in Computer-Aided Engineering from the University of Strathclyde. He is a programmer and
technology professional with over thirty publications in the fields of Engineering and Computer Science.
_____________________________________________________________________________________
 1

1. INTRODUCTION

Networks created by digital blockchains have now been around for nearly a decade. Digital contracts

and cryptocurrencies created on blockchains, for instance, result in a de facto network of users connected

in digital space that could interact with other users in exchanging instruments or engaging in transactions

on that network. Due to the fact that these networks are digital, online, as well as published on public

ledgers, the blockchain, data on the network is more readily available than most prior networks. This af-

fords us the ability to analyze these net-works in unprecedented ways, and possibly derive ma-cro-

mathematical relationships or models between entities on the network.

Nearly forty decades ago, Robert Metcalfe proposed a relationship for the value of a network in rela-

tionship to the size of the network (Metcalfe 2013). He stated that the value of the network is proportional

to the square of the nodes of the network (V ∝ N2) The relationship between the value of the network and

the size of the network is known as the network effect. Metcalfe’s Law has the form in Eqn. 1.

V(N) = kN 2 (1)

These relationships had been hitherto difficult to verify due to the fact that data needed for that verifi-

cation was difficult to obtain, as well as how to measure and define network value. However, several

works on social media networks recently have seemed to validate Metcalfe’s relationship when applied to

large social networks. They include Facebook (Metcalfe 2013, Van Hove 2014), and later Tencent (Zhang

et al. 2015), China’s largest social network. It is possible that the relationship between the value of a net-

work and the size of its nodes may depend on the type of network, and networks that can grow freely and

organically with little barrier or costs to join the network could follow Metcalfe’s relationship, while other

types of networks may not. In addition to proposing a value to network size relationship, Metcalfe also

proposed a relationship for the growth of the network size in relation to time, which he termed the netoid

function.

There have been other several models proposed for net-work growth (Swann 2002, Briscoe et al.

2006, Reed 1999). This includes models by Sarnoff (V ∝ N), Odlyzko (V ∝ N log N), and Reed (V ∝ 2N).
 2

However, Metcalfe’s model has been shown to be closest in approximating the growth on online networks,

and is presented here, in addition to a new proposed model. Three other models were also reviewed but

did not fit the data on any of the blockchain assets reviewed here, and thus were not explored further.

An objective of the current work is to determine if network of users of digital assets or other kinds of

assets connected via a blockchain, where those assets may be exchanged or transacted over the network,

have similar growth patterns or relationships, as was observed for social media networks.

For blockchains, there are several sub-networks in the system. The first is the network of end users

connected over the blockchain and holding assets that can be exchanged or transacted among each other

on the network. The second is the network of entities and computers that maintain the blockchain and

process transactions on the network. A third is the network of connected merchants on that network, pro-

vided the digital asset is a medium of exchange for merchant goods or services; or of developers and

companies developing products on the network, if the network allows the creation of third-party products

or instruments. It is likely that the value of the network will have a positive mathematical relationship to

each of these sub-networks. However, the current study is focused on deriving relationships on the net-

work effect as it relates to the number of end users on the network. It also examines the rate of growth in

time of some of the blockchain networks, and compares this to the netoid growth function. An alternative

model for both the growth function as well as the value of the network is also proposed.

2. METHODOLOGY

The blockchains selected were all cryptocurrencies since the price and capitalization of the networks

are available and published daily – in fact, every second. So are several other interesting facts, such as the

number of unique IP addresses participating on the network, the volume of transactions, and the number

of wallets relative to the user accounts on the network. The following are the parameters and assumptions

of this analysis.

(1) The value of the network is modeled by the price of the digital currency established by the net-

work. The price and capitalization have a direct relationship so either can be used without affect-
 3

ing the form of the model.

(2) The number of end users was selected as the number of unique addresses participating on the

network per day. This is in lieu of obtaining the actual number of active users on the network,

which is difficult to decipher due to the privacy of account ownership, and since a user can es-

tablish several accounts, or wallets, in participating on the network. The assumption is that given

a large number of participating users, the ratio of the number of actual users to the number of

unique address is roughly consistent.

(3) The actual curve of price on the network is assumed to consist of bubbles and bursts – noise re-

lated to medium- to short-term events that need to be filtered out in order to obtain the true mod-

el of growth and value of the network. This is slightly similar to filtering of the smaller turbulent

signals or length scales in fluid dynamics to derive the major length scales in a fluid flow. The

daily number of unique address was also filtered to remove the statistical noise from the data.

(4) The growth of the network is assumed to begin from the point when critical mass is reached. The

critical mass is a threshold number of users from which the network begins to grow or be-comes

viral. Below this number, the value of the network remains roughly stagnant and close to its ini-

tial starting point. The growth of the network is also expected to have a deceleration

2.1. Growth Model

This is a model for the growth of the network over time. The netoid function proposed Metcalfe Bris-

coe et al. 2006), is in the form of a function, with initial rapid exponential growth, and later as the net-

work reaches saturation, exponential deceleration with similar growth and deceleration rate. The netoid

takes the form of a sigmoid function as shown in Eqn. 2:



 = 
    (2)

where N, the total number of users, asymptotically goes to p as t tends to infinity. Also, tm is the center of

the sigmoid, where the number of users is assumed to be at peak growth rate and N = p / 2. Finally, v is

how fast the network is growing, also called its virality.

 4

None of the blockchains we examined has yet reached a point of stagnating value, so we are only able

to model the initial growth portion of the network. In addition, there no basis to expect that the parameter

v will also govern the deceleration rate of the network. Consequently, the growth models examined here

will focus on the initial growth section of the network following the time that critical mass has been

achieved.

2.2. Rationale for the Form of the Growth Rate

We next offer a rationale for the exponential relationship of the netoid. Due to the fact that the assets

we consider have very little actual use currently – they are not really used much as a transactional medi-

um for payment systems, nor even significantly as a payment rail – it may be reasonable to in-terpret their

adoption to be based on their perceived future potential. If that potential or value proposition is presented

to a selected number of people, the expectation is that the same ratio in that set would convert into acquir-

ing the assets. That group would then communicate that same value proposition to their own social net-

work of friends and acquaintances, and on. Therefore, the basic adoption process might not be very dis-

similar from that of a social network. So, if N is the number of people adopting the assets at time t, at time

t + t the number should be proportional to the number of people currently holding the assets. This can be

written as:

∂n / ∂t = vN (3)

That equation has solutions of the form:

N(t) = N0 evt (4)

2.3. A New Network Model for Blockchain Networks

A model for the value of the network in relation to the number of unique addresses participating in

the network utilizing an exponential function was also derived and compared with Metcalfe’s Law. The

new model has the following form:



 =   (5)

The values of C, λ, and m are determined by the fit to the equation. The filter applied to N is a 30- day
 5

average filter which takes the form:


 =

∑ 
  
 (6)


Note that h = 0 is the 30 day backward moving average and results in a lagging N in relation to the

value function. h was chosen to weight the filter evenly about the point at i for the exponential function,

with its effect shown in the close fit between the filtered and actual quantity in Fig. 1.

2.4. Data Sources

The data on daily closing price and unique participating addresses on the Bitcoin network, Ethereum

and Dash are available from the public blockchain, and were obtained from blockchain.info, etherscan.io,

and bitinfocharts.com.

3. RESULTS AND DISCUSSION

The results of applying the models are presented next.

3.1. Comparisons of the Network User Growth Models

Fig. 1 shows the comparison of the netoid function with actual data for Bitcoin going back nearly to

its advent in 2009.

Fig. 1. Comparison of the Netoid Function with Actual Data on Unique Addresses Participating in
Transactions per Day on the Bitcoin Network

The curves show good agreement with the live data on unique addresses per day transacting on the
 6

network. For the netoid function, the data fit as shown were obtained at a value of p = 1,500,000, v =

0.002, and tm = 3,200. The root mean square deviation (RMSD) of the model from the live data was com-

puted at 41,814, which is about 6% of the value of N at the end of the period that was analyzed.

Based on this model, tm will be reached in October 2017, and the network is already almost reaching

about half of p. In addition, while the prior agreement between the model and the growth in the number of

unique addresses participating on the network was good, there are known technical issues on the network

that remain unresolved at this point that may influence its future growth. This includes the maximum size

of a block in the blockchain that may limit the growth of the network in the future.

Figs. 2 and 3 show the respective netoid functions for the Ethereum and Dash networks.

Fig. 2. Comparison of the Netoid Function with Actual Data on Unique Addresses Participating in
Transactions per Day on the Ethereum Network

For the Ethereum network, the data fit was obtained at a value of p = 14,000,000, v = 0.01, and tm =

1,120. It is interesting to note that the growth rate of this network, as currently observed, is about five

times that of the Bitcoin network, and the number of unique addresses participating on the network has a

higher ceiling. In addition, the maximum growth time, tm, will be attained sometime in 2018 if the net-

work continues to follow the current model. The RMSD of the model compared to live data was comput-

ed at 14,333, which is about the current value of N at the end of the period analyzed.

For the Dash network, the data fit was obtained at a value of p = 1,600,000, v = 0.0024, and tm =
 7

3,200, which is surprisingly very similar to the Bitcoin network. However, as shown later in the analysis

of the point at which critical mass is reached, this network has just about reached this point, and so does

not provide sufficient information for analysis, as is available for the other two networks in this study.

Fig. 3. Comparison of the Netoid Function with Data on Participating Addresses in Transactions
Per Day on the Dash Network

3.2. Comparisons of the Network Value Model

Fig. 4(a) shows the comparison of Metcalfe’s Law, the proposed value to size function in the current

study, with actual data for bitcoin daily closing prices plotted over time. Fig. 4(b) shows the same plot for

the number of daily unique addresses. A scatter plot is used for price in Fig. 4(b), since the number of dai-

ly unique addresses participating on the network is not monotonically increasing. The figures show a

good fit between the models and the live data, except for the bubble in 2014. Fig. 4(b) is instructive in

demonstrating the trend between price and daily unique addresses, but since prices have a relationship

with their historical values, the plot based on time better illustrates the relationship between daily closing

price and the models, and is used for the rest of the comparisons presented.

The value of Metcalfe’s growth amplitude parameter k for the model is 4 × 10-9. For the proposed

function, 0.5, λ = 0.01, and C = 1. The RMSD of the models compared to the live data is 199 and 220 for
 8

Metcalfe’s Law, and the proposed function or about 8.7% and 9.6%, respectively, compared to the closing

price at the last data point.

Fig. 4. Comparison of Metcalfe’s Law and Value-Size Relationship

(4a) Bitcoin daily prices plotted over time

(4b) Bitcoin daily plotted for daily unique addresses

For the Ethereum network, with similar layout of Fig. 5, the value of Metcalfe’s growth amplitude pa-

rameter k for the model was 11 × 10-9. For the new function, m = 0.5, λ = 0.011, and C = 3. The RMSDs

of the models compared to the live data are 14.7 and 13.8, respectively, for Metcalfe’s Law and the pro-

posed function. This is approximately 4.3% and 4.1%, respectively, compared to the closing price at the

last data point.

For the Dash network, shown in Fig. 6, the value of Metcalfe’s growth amplitude parameter k for the
 9

model was 3 × 10-7. For the proposed function, m = 0.5, λ = 0.033, and C = 1. The RMSD of the models

compared to the live data are 19.3 and 18.7, respectively, for Metcalfe’s Law and the proposed function,

or about 16% and 15% compared to the closing price at the last data point. The agreement is surprisingly

good for the networks analyzed.

Fig. 5. Comparison of Metcalfe’s Law and the Value-Size Relationship with Daily Ether Values on
the Ethereum Network

Fig. 6. Comparison of Metcalfe’s Law and the Value-Size Relationship with Daily Dash Prices on
the Dash Network

Based on the data, since m = 0.5 in all the cases observed, the form of the proposed model for the

growth of the network as a function of the number of unique addresses participating actively on the net-

work is:
 10


 =  √ =   (7)

In other words, the observed growth rate is exponentially proportional to the root of the number of ac-

tive users. The form of the proposed model allows some interesting value-based comparisons of each

network beyond what Metcalfe’s Law provides. Here, γ = λ2 , is a measure of the exponential multiplier

for the active users, while C is a linear multiplier or linear growth factor. For instance, a network imple-

menting a payment medium may have a higher growth multiplier if it is accepted by more merchants than

another payment solution.

Another interesting observation is that the form of the proposed model does not result in a pre-

ordained exponent in the manner that N2 does. In short, it does not guarantee aarge exponential growth

with N, which has been a criticism of the Metcalfe’s Law (Briscoe et al. 2006). In fact, the growth rate γN

may be greater or less than that of N2, depending on the type or the intrinsic dynamics of the network.

This allows the model to be utilized even as the network reaches maturity. For instance, a value of zero

results in a network whose value no longer grows exponentially as it adds new users.

3.3. Analysis of Critical Mass

Figs. 1 and 4 showed that the growth and value of the network remained flat from inception up until a

point in late 2013 when the growth in the number of users became exponential (Ball 2014). It also is use-

ful to determine whether some parameter or model can be applied to determine roughly at what point a

blockchain network reaches the point at which its exponential growth begins. Fig. 7 shows a plot of the

first derivative of growth over time for the unique daily addresses that participate in the Bitcoin and

Ethereum networks.

We expect a network that has not reached critical mass will not have a rapidly growing user base. The

first derivative of N for time should be close to 0 and nearly constant in the early stage. Based on this ob-

servation, it is possible that, once the derivative grows to a critical number, and the number of daily

unique participating addresses also reaches a critical threshold, the network will begin to show trends as-

sociated with rapid growth. These two conditions combined can be expressed as follows:
 11

Fig. 7. First Derivative of Daily Unique Address Participation in Networks with Respect to Time

(7a) For the Bitcoin network

(7b) For the Ethereum network



!"

# 
,
!"

(8)


More work needs to be done to confirm these conditions, using more data as it becomes available.

However, based on these criteria, for the above two networks, the critical parameters appear to be about

100, and 20,000, respectively.

3.4. Value Bubbles

Another observation that can be gleaned from an analysis of the value of network based on modeling

its growth with user adoption is that value bubbles can possibly be identified. In Fig. 1, for instance, an
 12

increase in the value of the network can be observed from late 2013 to early 2014. That growth was not

accompanied by anything close to a commensurate increase in the daily number of unique addresses par-

ticipating in the network, and was a deviation from the underlying network models presented here. How-

ever, that value deviation turned out to be temporary, and the value snapped back to the model or closer to

a value representative of the number of users participating on the network. A network growth model anal-

ysis can allow these types of value bubbles to be spotted more easily than without such models.

3.5. Long Term Persistence of the Network

In as much as the value of the network correlates with the number of users actively participating in it,

that value can also prove to be as fleeting as the ease with which those users can move to a different net-

work or cease to participate. The idea behind some of the assets designated for use on the networks stud-

ied here is one of fungibility. That fungibility also means that users can create new addresses on different

networks and move their assets there easily, or simply just pull their assets out. In short, the ease with

which users can move from the blockchain networks of the types studied here exceeds that of networks

such as social media, where the user may have cherished items including pictures, conversations, social

contacts, and other historical items that may not be as easy to move.

As a result, without some context, care should be taken in considering the value of these networks at a

point in time to be on par with other different types of networks that may have exhibited similar growth in

relation to their number of users in the past. Any form of lock-in of the value of the network to any extent

may turn out to be more related to some of its other sub-networks mentioned earlier. This includes how

many businesses or merchants are participating in the network, how many developers have invested time

and resources in creating products on the network that are of value to users, and how much hardware and

nodes that are not easily transferred and adapted elsewhere are invested in the network.

4. CONCLUSION

An analysis of selected blockchain networks was presented, and it was demonstrated that the growth

in the value of the network was related to the number of unique addresses participating actively on the
 13

network. The number of unique addresses participating daily was suggested as a relative measure of the

number of active users on the networks. Metcalfe’s Law of network value, which correlates the value of

the network with the square of its number of active users, was shown to model the networks well. A new

model was also proposed based on an exponential of the root of the number of active users. The proposed

model was shown to also capture well the trend between the value and size of the users participating on

the network. Conditions for determining the point at which a network has reached critical mass were also

suggested.

Finally, the possibility of spotting value bubbles based on network growth model analysis was sug-

gested. Such bubbles can be spotted where the value of the network begins to grow rapidly unaccompa-

nied by accompanying growth in the number of users participating in the network or any other develop-

ment that could result in enhanced value to anyone or user. In addition, the potential ease with which us-

ers participating on the network could leave the network and other measures of network strength that

could indicate resilience of the network’s value was discussed.

REFERENCES
Ball, P. 2004. Critical Mass: How One Thing Leads to Another. Farrar, Strauss and Giroux, New York.
Briscoe, B., Odlyzko, A., Tilly, B. 2006. Metcalfe’s Law Is Wrong. IEEE Spectrum, July 1.
Metcalfe, B. 2013. Metcalfe’s Law after 40 years of Ethernet. IEEE Computer 46, 12, 26-31.
Reed, D.P. 1999. That sneaky exponential: Beyond Metcalfe’s Law to the power of community building.
Context Magazine 2, 1.
Swann, G.M. 2002. The functional form of network effect. Information Economics and Policy, 14, 3,
417-429.
Van Hove, L. 2014. Metcalfe’s Law: Not so wrong after all. Netnomics: Economic Research and Elec-
tronic Networking 15, 1, 1-8.
Zhang, X.Z., Liu, J.J., Xu, Z.W. 2015. Tencent and Facebood data validate Metcalfe’s Law. Journal of
Computer Science and Technology, 30, 2, 246-251.