Improvements on the Network Value to Transactions (NVT) Ratio & Introducing Network Value/Transactions to Growth (NVTG) to Value Crypto

This article is going to be the first in a series of developing fundamental ways to value crypto. If you haven’t checked out my previous articles summarizing the valuation space so far, you can view them here: Part 1 (Absolute Valuation) and Part 2 (Relative Valuation). The four main ways we have to value crypto right now are from its use as a currency, stock, product, or network. This will address valuation from the network perspective.


One of the most promising aspects underlying blockchain technology is the potential for us to be able to interact across a borderless, unrestrained network. Whatever this network is used for (identity verification, security, data storage, payments, etc.), the construction and implementation of this system is highly valuable. Moving forward, as blockchain technology continues to improve and expand across different lines of business, one of the most important metrics to describe its success is its network value.

How do we do this?

We need to design a fundamental metric to accurately capture the utility of a digital asset’s network. We can look at daily transaction volume in the form of on-chain transaction volume to get an idea of how “useful” the asset is. However, this by itself is not enough to create a relative valuation model.

In traditional finance, we use metrics like Price-to-Earnings (P/E) to measure how “cheap” or “expensive” an asset is — i.e. how much we must pay to get $1 of its earnings. The crypto analogy to this would be a ratio called Network Value to Transactions (NVT), initially popularized by the Coinmetrics team.

Original NVT Ratio

Network Value, measured by market capitalization, replaces price and daily transaction volume replaces earnings as a measure of how much utility value the asset provides. Essentially, this metric is measuring how valuable something is for how much it is currently being used. However, as described in this great article by Dmitry Kalichkin, the original NVT ratio (while an amazing theory) is not predictive or descriptive. You should check out his article but essentially what Dmitry and Cryptolab Capital found after experimenting is that the following NVT ratio works much better.

Kalichkin’s NVT

Referred to as the NVT Signal by Willy Woo, the original proponent of NVT, this can be used as a much more responsive trading signal. I won’t go into a proof of it’s explanatory and predictive power here, but you can see that in Dmitry’s article and also on Willy’s blog. This is due to smoothing the transaction volume over a longer period (the previous formula did it over 28 days instead) which gets rid of irregularities and noise distracting from the fundamental, longer term value of the network.

Proposed Changes & Additions to NVT

Kalichkin’s NVT is a great way to transform a fundamental metric into a dynamic trading signal. Analysis that we have conducted for bitcoin leads us to believe that two main improvements can be made:

1) Changing the moving average from simple to exponential

Simple moving averages across a long-time span (such as 90-days) are not as responsive to short-term movements as their exponential counterparts. The simple moving average used in the denominator of Kalichkin’s NVT can be calculated by the following formula, where TV is the daily transaction volume and n is the timeframe:

Simple Moving Average Method

As you can see, all historical values are weighted equally when smoothing the current value. In a fast-paced space like crypto, we need to weight more heavily on the most recent data to get rid of time lag. This is especially relevant since we are trying to get an accurate representation of fundamental, intrinsic value by using a long moving average. To get the best of both worlds (fundamentals and a trading indicator), a 90-day EMA is most effective. This is calculated as follows:

Exponential Moving Average

The results of this change are shown below for bitcoin. Many time frames were tested and similar to Kalichkin’s findings, we found that a 90-day period was the most compelling. The EMA tracks the daily price almost exactly up until the past 30 days, where it is much higher indicating some sort of overpricing (at least on a historical basis).

Revised NVT Formula
EMA NVT vs. Price of Bitcoin
Comparison of NVT methods

We can see that the EMA NVT is less than Kalichkin’s NVT (indicating a more attractive investment) until the price starts falling. At this point, the exponential metric realizes this and quickly becomes larger than Kalichkin’s NVT around January 2nd, 2018 correctly indicating that bitcoin is becoming more overpriced. It appears that the EMA NVT is like Kalichkin’s NVT in all ways except for being more receptive to short term changes which makes it a better functioning trading indicator. Additionally, there are applications in using the crossover in these two NVT lines to determine a shift in momentum.

2) Using On-Chain Volume Excluding Long Transaction Chains

There are some concerns out there that on-chain transaction volume is not the most representative measure for determining “value-added” transactions. The Coinmetrics team points out the two biggest issues in this great piece that you should really check out:

1. It’s hard to tell which outputs are genuine

2. A large portion of bitcoin transaction volume comes from coin mixing, or exchanges/wallets just moving money around

To solve this issue, created a revised metric to remove transaction chains > 100, which are typically (but not always) caused by coin mixing and attempts to manipulate transaction volume. In light of the recent OKEX and Huobi scandals, this is very important to maintain integrity in the markets. We found that there is no significant difference in using this metric so far, but if available (currently only works for bitcoin) it should be used. NVT values are more sensitive since the denominator of transaction volumes are lower.

Using On-Chain Volume — Long Chains

Introducing Network Value/Transactions to Growth (NVTG)

This is where it gets really interesting. In finance, the two most common metrics we use are Price-to-Earnings (P/E ratio) and Price-to-Earnings/Growth (PEG ratio) to value stocks. The P/E ratio is used mostly to value stable, good long-term bets. It falls apart when applied to stocks with huge growth potential — for example, a company developing the cure for cancer would have almost no earnings initially, resulting in a very high P/E. However, the growth potential of this company would be undeniable and this needs to be factored in to accurately judge the company. The reason the PEG ratio exists is to account for the differences in growth potential that each company would have. This has great applications to the dynamic crypto space.

To analyze growth, just looking at transaction volume is not enough. We can take inspiration from Metcalfe’s Law, an equation that states that the value of a network is proportional to the square of the number of connected users of the system. In valuing digital assets, we would want to define the N users as the number of unique daily addresses. Therefore, the growth in network value would be equal to the derivative of the law being observed. This has been proven to work by itself in valuing crypto assets, by Ken Alabi and Clearblocks. An alternative equation to Metcalfe’s Law, Zipf’s Law will also be investigated after Andrew Odlyzko conducted a very convincing study.

Fundamental equations of network value and growth

Using these equations, two variables of Network Value/Transactions to Growth (NVTG) was created:

NVTG using Metcalfe Growth
NVTG using Zipf Growth

We found that NVTGZ was not really a useful trading indicator. We hypothesize that this is because the crypto markets are still very new and have not reached the “saturation” stage that Zipf’s Law indicates with the logarithmic scale. NVTGM captures the growth period that we are in right now, but perhaps later on we will see that Zipf’s Law is a more accurate characterization. Smoothing using a 90-day EMA worked the best in terms of predictive and descriptive power out of all the methods. The following figure, with arbitrarily selected “overpriced” and “underpriced” red and green lines respectively, shows these results.

Using NVTGM as a trading indicator

Whenever the blue NVTG line crosses below the green line is when bitcoin is relatively underpriced and when it crosses above the red line is when bitcoin is overpriced. We can see that these arbitrary thresholds have served well over the past 2 years, correcting identifying the three times it was overpriced and five times it was underpriced. However, we can see that it doesn’t work as well during the last 50 days where it said that bitcoin was underpriced yet the price kept falling. Some more work can be done to create optimal thresholds given a certain timeframe (trying to optimize for a 2-year time period is too optimistic in this space), like 1-month or 3-months.


Network Value is a very important metric that shows how valuable a cryptocurrency is. In this piece, we made changes to Kalichkin’s NVT (a.k.a. NVT signal) to create a more responsive trading indicator that still retains an accurate valuation of the fundamental network value. We also introduced a brand-new metric to factor in the growth in network value derived from Metcalfe’s Law. This metric, which we call NVTG, was inspired by the uses of the PEG ratio in traditional finance.
Crypto corollaries to traditional equity valuation

NVT neglects the network value created by additional users, assuming that transaction volume would reflect this, but the NVTG factors this in to compensate for network growth more accurately. In practice, these metrics could be used together and with different time periods to create a more comprehensive understanding of network signals.

So… what’s next?

This analysis was only performed for bitcoin. Keep and eye out for a more comprehensive analysis of these metrics for other cryptocurrencies. Additionally, a lot of the thresholds are subjective and creating more of a structure around what levels truly indicate mispricing by asset is something that could be done. And if you have any comments, questions, or any suggestions on what to look at next, let me know!