Nominated Proof of Stake Empirical Analysis

Abstract

Consensus protocols underpin the security and correctness of decentralized blockchain systems by enabling mutually untrusted participants to agree on a shared state. Early blockchain networks relied on Proof-of-Work (PoW), which achieves strong security by making participation computationally expensive, but its high energy consumption has driven the transition toward Proof-of-Stake (PoS) and its variants. Nominated Proof-of-Stake (NPoS) addresses PoS centralization tendencies by separating stake from block production, allowing participants (known as nominators) to delegate their staked tokens to validators, aiming to enable broader, fairer, and more decentralized participation in consensus. This thesis addresses the gap between NPoS’s design intentions and its realworld effectiveness by using Polkadot as a case study: we employ multi-year on-chain data and a graph-based methodology that models the evolving relationships between nominators and validators, and introduce metrics to evaluate centralization, fairness, and inclusiveness in NPoS systems. The analysis reveals persistent structural concentration. Validator turnover is limited, with around 90% of validators remaining active across consecutive eras, and nomination patterns are highly persistent, with fewer than 20 nominators controlling up to one-third of the active validator set. Regarding fairness, while validator rewards converge to near-equality, nominator rewards remain highly unequal as an inherent consequence of stake-proportional allocation. Regarding inclusiveness, participation remains constrained: a non-trivial fraction of active validators charge 100% commission, retaining all staking rewards and distributing nothing to their nominators, which constitutes a structural barrier to meaningful economic participation for the nominators backing them. Overall, these findings indicate that NPoS still exhibits concentration dynamics that limit its effectiveness in achieving decentralization and inclusiveness.