Skip to main content

The Delegation Problem

This doc is the shared baseline for all proposal reviews.

Goals (primary)​

  1. Increase net-new delegated stake (not just reshuffling existing LPT).
  2. Increase the number of independent stake decision-makers in meaningful stake brackets — especially 1k–10k and 10k+.
  3. Improve retention (reduce fast churn/unbond/withdraw patterns).
  4. Improve decentralization (avoid concentrating new stake to a small set of orchestrators/delegates).

Note: “small delegators” matter for community participation, but current on-chain distributions show they are economically negligible by stake, so a strategy that only optimizes for “more small addresses” will not move security metrics.

Constraints / realities​

  • Arbitrum environment: low fees help, but sybil is still cheap.
  • Delegation incentives are sybil-sensitive when rewards scale by “# of accounts”.
  • Many proposals increase “delegation count” while decreasing security if they over-concentrate stake or can be farmed.
  • A material share of LPT rewards appears to be withdrawn and bridged/transferred out (see /research/reflexivity-and-yield-extraction), so “more LPT incentives” can translate into more structural sell pressure unless retention-gated or tied to real fee growth.

Canonical metrics (must be reported for any solution)​

  • New delegators (first-time bonders): daily/weekly/monthly
  • Net stake change: net bond − unbond − withdraw
  • Rewards issuance vs withdrawal: LPT rewards claimed (see /research/rewards-withdraw-timeseries) vs withdrawn (upper-bound + conservative proxy)
  • Bracket distribution: active wallet counts + bonded LPT in 1k–10k and 10k+ (and how that changes)
  • Retention: % of new delegators still bonded at 30/90/180 days
  • Concentration:
    • delegators (top-N share of bonded stake)
    • delegates/orchestrators (top-N share, Nakamoto 33/50)
  • Sybil risk proxies: funding clustering, common senders, fast-exit cohorts

Common failure mode (call it out explicitly)​

If a proposal claims “we will increase small delegators” but its mechanism is “pay more per address for smaller balances”, it is sybilable by design unless it has a uniqueness primitive (identity, proof, or a strong economic cost to splitting).