How we score · version 1.0
How the Decentralization Index works
No single number captures decentralization. A network can have thousands of nodes but only three companies producing blocks, or a widely spread validator set but one foundation that can change the rules. So the index combines seven measures, scores each one from 0 to 10 with fixed thresholds, and takes a weighted average.
The seven measures
| Measure | Weight | What it captures |
|---|---|---|
| Nakamoto coefficient | 25% | How few block producers could together halt or censor the network |
| Full nodes | 15% | How many independent computers check every rule |
| Client diversity | 10% | Whether one software bug could take down most of the network |
| Hardware to run a node | 10% | Whether ordinary people can verify the network themselves |
| Token concentration | 15% | How much of the supply sits in the largest addresses |
| Insider allocation at launch | 15% | How much went to founders, investors and foundations |
| Control over upgrades | 10% | Whether one party can change the rules or pause the network |
How each measure is scored
Scores rise or fall in a straight line between the points in each table, and stay flat beyond the first and last points.
Nakamoto coefficient
| Coefficient | 1 | 3 | 7 | 20 | 50 or more |
|---|---|---|---|---|---|
| Score | 0 | 3 | 5 | 7.5 | 10 |
Full nodes
| Reachable full nodes | 0 | 100 | 1,000 | 10,000 | 20,000 or more |
|---|---|---|---|---|---|
| Score | 0 | 2 | 5 | 8 | 10 |
Client diversity, from the share of nodes running the most popular client software
| Largest client share | 33% or less | 50% | 66% | 90% or more |
|---|---|---|---|---|
| Score | 10 | 6 | 3 | 0 |
Hardware to run a node
| Needs | A typical home computer | A powerful desktop | Data-center servers |
|---|---|---|---|
| Score | 10 | 6 | 2 |
Token concentration, from the share of supply held by the 100 largest addresses
| Top 100 share | 10% or less | 20% | 40% | 60% | 80% or more |
|---|---|---|---|---|---|
| Score | 10 | 7.5 | 5 | 2.5 | 0 |
Insider allocation at launch uses its own score. See how we score insider allocation.
Control over upgrades
| Who can change the rules | Score |
|---|---|
| No single party: node operators each choose whether to run new software | 10 |
| A core team or foundation coordinates upgrades, but operators must choose to adopt them | 6 |
| Admin keys or a small group can upgrade or pause the network on their own | 1 |
Combining the measures
The index is the weighted average of the measures we have data for. We only publish it when we have the Nakamoto coefficient plus all but two of the measures that apply (and never fewer than four). The profile page shows exactly which measures were included.
Networks without a token of their own, such as many layer 2s, have no launch allocation or token concentration to measure. Those two measures don’t apply, and the remaining weights are used.
Limits we’re open about
- For proof-of-work networks, the Nakamoto coefficient counts mining pools. Individual miners can switch pools, so this overstates concentration somewhat.
- Node counts come from public crawlers that only see reachable nodes. The true number is usually higher.
- Large addresses are often exchanges holding many people’s coins. We exclude known exchange, bridge and contract addresses where the data allows, and say so in the notes.
- Every figure is a snapshot. We re-measure at least every three months, and each figure shows its date.
Where else it’s used
The index is also the “Decentralization” criterion in our network scores, which feed the setup finder’s network recommendations.
See these scores in action: compare networks side by side.