The Nakamoto Coefficient of Polkadot is At Most 25

Ecosystem
1

Recently I went down the rabbit hole to calculate Polkadot’s Nakamoto Coefficient number.

The reliable approach

The methodology is simple:

  • Validators are grouped by their on-chain identity (sub-identities as well as obvious common names).
  • Each anonymous validator is considered an independent one.

We then check for how many operators we need to reach 33% of bonded stakes.

The data:

  • 374 validators are anonymous.
  • The rest are validators with identities. Grouped by their sub-identities, there are 94 independent operators in total.

With this data, the Nakamoto Coefficient of Polkadot is 25.

Anonymous clustering

You may have seen above that we assume each anonymous validator is independent, this means that 25 is still just the upper bound.

To dig into the anonymous validators, we then must use some clustering approach. My clustering approach, grouping validators by their transaction counter-parties, gives me the Nakamoto Coefficient of Polkadot at 14.

The often cited value of 178 is wrong

So there you have it. If we use only hard, cold, reliable data (the reliable approach), then we have the Nakamoto Coefficient of Polkadot at 25. If we additionally group anonymous validators (anonymous clustering), then the number will likely be lower.

In any case, the often cited value of 178 is almost certainly wrong. Citing Polkadot’s NC as 178 is the same as citing Ethereum’s NC as 300000.

Verify my data

If you want to verify my data, and have been an old guy like me, to avoid unnecessary detours for you, just two things to take note:

  • The identity data is now on the “People Parachain”, not on relay chain.
  • Four operators do not (completely) use sub-identities. Just pattern match their names: Binance, Figment, RockX, ChainSafe.
4 Likes
2

An also likely interesting reading for you is Validators running multiple identities and flipping commissions in same day - #51 by Megadot

The Nakamoto Coefficient can still be lower if that fact is considered.

1 Like
3

Where did 178 come from?

4

Flipping commissions and misrepresenting identities is poor sportsmanship and generally a “dick move” but does that directly represent a threat to the security of the network? So long as the validators are building valid blocks and assuming the risk of slashing their stake if they misbehave, it shouldn’t matter too much.

Also, keep in mind that the validators are grouped randomly, so assuring all their fellow validators are also interested in collusion of block production may still be statistically improbable at 25.

Still worthwhile to increase independent, individual validators, though.

5

Flipping commissions and misrepresenting identities are not yet considered. If they are considered, we’ll get an even lower NC.

I don’t think the “grouping” you mentioned makes any difference – it’s how parachains are validated. If the relay chain is already compromised (>33%), everything stops working already.