Nakamoto Coefficient

The Nakamoto Coefficient shows how decentralized a blockchain network is by counting the fewest independent groups needed to take control or disrupt the system. Simply put, it tells you how many people would need to work together to change or stop the chain.

The term became popular after a 2017 essay by Balaji Srinivasan and Leland Lee, who wanted to make the idea of “distributed power” measurable. The name refers to Satoshi Nakamoto, the creator of Bitcoin. People use the term because it offers a simple way to discuss decentralization.

How the coefficient is calculated

To find the Nakamoto Coefficient, choose a part of the system to measure, list the main players by how much power they have, and add them up from largest to smallest until their total power passes a set threshold. For proof-of-work chains, this is usually 51 percent of hashing power. For proof-of-stake chains, the threshold might be one third or 50 percent, depending on the rules. The number of players needed to reach that point is the coefficient for that part of the system.

Different dimensions to measure

Decentralization has many sides, so the coefficient can be used for different parts of a system. Common areas include mining or block production, client software, code development, exchanges, node count, and token ownership. Each area shows a different type of concentration, and a network might be decentralized in one area but centralized in another. Analysts usually report the lowest number to show the weakest spot.

Who uses it and why

Developers, investors, stakers, and researchers use the Nakamoto Coefficient to quickly check for centralization risks. Protocol teams look at it to find where power is gathering and may adjust incentives or rules. Investors include it in risk checks, and people who delegate stake might choose smaller validators to help raise the number. The metric is useful, but most people see it as just one part of a bigger picture.

Limitations and criticisms

The Nakamoto Coefficient gives a helpful snapshot but leaves out some details. It does not show where actors are located, if one group runs many nodes, or if different parts rely on the same provider. It can also be unclear what counts as a single “entity.” The chosen threshold matters, since some systems are at risk with a one-third majority while others use 50 percent. Also, short time frames and measurement errors can make the number change a lot, so it is better to look at trends or ranges instead of just one value.