Network

How concentrated each chain is

A long validator list is not decentralisation. What matters is how few parties it would take to stop the chain — and on one of these three, a set of 898,811 “validators” comes down to seven.

How few it would take

holders needed to reach a third of the stake

Solana

of: 682 validators
largest holds: 3.8%
behaves like: 103 equal

where the stake sits

a third

One row is one voting validator. Delinquent validators are excluded: they are not currently participating, so they neither secure the chain nor could halt it.

BSC

of: 45 validators
largest holds: 6.2%
behaves like: 27 equal

where the stake sits

a third

One row is one elected validator. The registered set is larger, but only the elected 45 produce blocks, and only they could stop.

Ethereum

of: 43,378 withdrawal addresses
largest holds: 25.0%
behaves like: 15 equal

where the stake sits

a third

One row is one withdrawal address, not one validator. A validator is 32 ETH and a single pool runs hundreds of thousands of them, so counting validators would report this as the most decentralised chain here by a factor of ten thousand — when it is the one where a single party controls the largest share.

The shape of each distribution

Solana682 validators
BSC45 validators
Ethereum43,378 withdrawal addresses

Each marker is a measured point: what the top one, five, ten, twenty, fifty and hundred hold between them, and where the whole set reaches 100%. A curve that climbs steeply and flattens early is a chain whose stake sits with a handful of holders. Buckets larger than the set itself are left off rather than drawn at 100%, so no segment claims a measurement it does not have.

The horizontal axis is logarithmic, because these three sets differ by three orders of magnitude and on a linear axis the smallest would be a vertical line at the origin.

The distribution

Stake concentration by chain
Chain	Counting	Nakamoto	Holders	Largest	Top 10	Top 20	Effective	Gini
Ethereum	withdrawal address	3	43,378	25.01%	38.4%	42.7%	14.6	0.950
BSC	validator	7	45	6.23%	46.8%	82.9%	26.8	0.448
Solana	validator	19	682	3.84%	24.0%	35.0%	102.5	0.723

Take it awayCSV JSON

Nothing in this table is comparable across chains without reading what each one counts. The column exists for that reason and is never omitted.

How to read these numbers

Nakamoto coefficient: The fewest holders whose combined stake reaches a third — the share at which a party can halt a chain by refusing to participate. It is the single most useful figure here, and the one a validator count is most often mistaken for.
Effective holders: How many equally-sized holders would produce the same concentration. It is the readable form of the Herfindahl index: a chain with 41,807 addresses that is as concentrated as 19 equal ones is 41,807 in name and 19 in practice.
Largest, top 10, top 20: Cumulative share held by the biggest one, ten and twenty. Where the Nakamoto coefficient answers “how few could stop it”, these answer “how much sits with the few”, which is not quite the same question.
Gini: Inequality across the whole distribution, from 0 when every holder is equal to 1 when one holds everything. A chain can have a decent Nakamoto coefficient and a very unequal tail, or the reverse — which is why both are here.

None of this is a judgement about any operator. Concentration is a property of where stake has gone, and every delegator who chose the largest validator was making an individually reasonable decision. It is precisely because that decision is reasonable that the aggregate is worth measuring — and why the staking score prices it in rather than recommending the top of the list to everybody. Staking yields

The Nakamoto coefficient is the fewest holders whose combined stake reaches a third — the share at which a party can halt a chain by refusing to participate. Effective holders is the number of equally-sized holders that would produce the same concentration, which is the readable form of the Herfindahl index beside it. Nothing here is comparable across chains without reading the unit each one counts.