Measuring Proof-of-Stake Networks

etherfuse’s mission is to further decentralize blockchain infrastructure by lowering ownership barriers and empowering owners to leverage their assets to improve their lives and communities. Our mission is ambitious. To succeed, we're required to create software, models, infrastructure, and regulatory policy which enable anyone to participate in the future. We’re primarily focused on proof-of-stake networks and to start we need to be able to measure the decentralization of these networks.

How we can measure our efforts? A lot of work has gone into the decentralization of blockchain tokens and proof-of-work networks. Many have overlooked the importance of decentralizing the infrastructure that powers these technologies — especially when considering proof of stake networks. Fortunately, we get to stand on the back of a giant and use the system of measurement proposed by Balaji. If you're unfamiliar with his work, take some time to read his paper before going forward — as it's a prerequisite to our work. We use the same approach. However we propose the addition of five new subsystems to better understand the decentralization of the infrastructure that power proof-of-stake networks.

By proposing these subsystems, we attempt to quantify the decentralization of ownership, political, algorithmic, and corporate influences. As an example, we will use the Solana network. We will only focus on calculating the Nakamoto Coefficient for each subsystem and will defer the measurement of Gini Coefficient for a later post.

We have created two charts for each element used to measure a subsystem. The chart on the left will be the total count of that element in the network and the chart on the right will be the cumulative percentage of that element compared to the cumulative percentage of total validators in the Solana network. The only exception from this pattern will be when we measure algorithmic influence. In that case we’ll use cumulative stake in the network compared to cumulative stake in a stake pool.

For example when looking at the subsystem political influence we’ll use country as an element. In the above graph on the left is all validators in each country sorted in ascending order and the graph on the right is the cumulative percent of countries compared to the cumulative percentage of all validators. This helps us calculate a Nakamoto Coefficient for each element being measured.

Our data primarily comes from the APIs by Stakewiz and Solana Compass  — many thanks to them for their resources.

Terms

• Super Minority - The smallest amount of validators needed — which hold 33% of all stake — to collude in-order to compromise the network.
• Nakamoto Coefficient - The smallest measurement of the element being measured needed to compromise the network.

Proposed Subsystems

• Political - (by country)
• Regional (by Region)
• Corporate (by ASN and IP)
• Ownership (by wallet)
• Algorithmic (by stake pool)

Political Influence

We're not going to attempt to aggregate countries' politics but rather propose to present them as independent political entities. One may suggest that Canada and the United States could aggregate into a similar political ideology —a democracy. In the crypto context, they have proven that they have different ideas — so we define them to be separate political entities.

The graph on the left visualizes our intuition that most of the infrastructure resides in the West. The right highlights what may not be so intuitive — a super minority operates in Germany and the US.

The cumulative percentage of validators shows that the network has a Nakamoto Coefficient of two when measuring the subsystem by political influence. Germany and the US contain enough validators that could conspire to control the network.

Regional Influence

We could intuit that most validators are in the West. The data confirms our intuition when looking at the count of validators by Region.

When looking at the Regional Subsystem, the network has a Nakamoto Coefficient of one — Europe. A problem to solve is the lack of any validators in the Caribbean, Oceania, and Central America.

Corporate Influence

To measure the decentralization of corporate influence, we use two subsystems: IP organization and ASN. The danger of corporate power is similar to that of political forces. A centralized governance structure dictates policy that could use its leverage to censor.  

Centralization around the private networks of a few represents what is currently the most significant risk to the network. The sheer dominance of three or four companies is enough to have a super minority. More surprisingly was how little influence Amazon, GCP, and Azure had relative to their Web2 dominance.

The Nakamoto coefficient for IP and ASN is five and three, respectively. However, two of those five IP networks have one company governing both. We propose measuring this subsystem with both elements as they are a less informative measurements alone.

Ownership Influence

The most challenging part of measuring ownership influence is recognizing that there is currently not a good way to do it. It’s possible to have one person manage multiple validators and likely that they do. Detecting this is difficult but identifying this risk is the first step to improving how we measure the influence of ownership.

We propose this subsystem as a measurement to further highlight its imperfections. However, amongst other subsystems, it's a piece of data that helps us measure decentralization. Individually, it's not that valuable as it’s currently assumed to be.

Decentralization by the owner has a Nakamoto coefficient of 25 - 27, depending on how you slice the data. This measurement is the primary number referenced, when discussing decentralization in proof-of-stake networks — so it should come as no surprise. What should come as a surprise is how much we trust the unit of ownership in isolation to measure the network.

Algorithmic Influence

Stake Pools act as an intermediary between the person staking and a validator. They can be a massive catalyst for the decentralization of infrastructure on proof-of-stake networks. But, it's not without its costs. Marinade Finance is the dominant Stake Pool on the Solana network — having a centralized-algorithmic influence. They control three times or more the stake than the nearest competitor Lido.

The Nakamoto Coefficient of algorithmic decentralization is one. If you are deciding to invest in creating a validator, you would need to conform to the criteria the pool uses to delegate stake to validators. As powerful as the algorithm that powers who gets heard on Twitter, these algorithms are more powerful in determining who gets to participate in the future — that is why we propose it as another subsystem. We need to measure, watch, and influence these algorithms

The barriers to participate in the growth of the networks that will power the future is prohibitively high. We are building tools, services, models, policies, and networks to eliminate those barriers and improve the decentralization of proof-of-stake networks.

We will use these subsystems to measure our actions, guide our decisions, and analyze our impact — you should too.

If you like, hate, or are in-different towards our approach: stake with us, subscribe to us, or follow us on Twitter.