Transparency — stake.one

1. What we are measuring

The experiment measures the realized net spread between two staking strategies that share every variable except the staking mechanism itself:

Direct arm (passive). The demo wallet delegates 4,000 ONE proportionally across the same validator set the vault uses, then does nothing. Rewards accrue as pending claims on-chain and are never claimed or re-delegated for the duration of the experiment. This models the simplest possible reference point — “just delegate and leave it.”
Vault arm. The same demo wallet deposits 4,000 ONE into the vault. The vault routes to the same validators per the same weights and auto-compounds via permissionless compound() calls. The vault charges a 1% protocol fee on realized rewards.

The visible delta is the compounding lift minus the 1% protocol fee, at the prevailing network APR. Compounding lift is sub-linear in APR: at low rates the protocol fee roughly cancels it; at higher rates or longer windows the lift dominates and the vault pulls ahead. The realized spread is the empirical answer to that trade-off, with no cherry-picking.

2. Experiment design

Size per arm: 4,000 ONE
Total demo capital: 8,000 ONE
Window: 90 days fixed (no rolling)
Compounding cadence: Vault: auto ~12h · Direct: never
Validator basket: IPH 50 / EasyNode 15 / V1 10 / Fortune 10 / KRATOS 10 / Mintbes 5
Re-mirror SLA: 24h on vault reweights
Headline metric: Cumulative net ONE delta
Sampling: Daily snapshot, not real-time

Direct-arm proportional allocation rounds to whole ONE: IPH 2,000 / EasyNode 600 / V1 400 / Fortune 400 / KRATOS 400 / Mintbes 200 (sum 4,000 ONE).

3. What we'd expect

With a passive direct arm, the realized spread (vault − direct) decomposes into two competing forces:

Compounding lift (vault gain). The vault re-stakes harvested rewards on each compound. The longer the window and the higher the APR, the larger the compounding bonus over a wallet that lets rewards sit unclaimed.
Protocol fee (vault drag). 1% of realized rewards on the vault arm flows to the protocol (split 90/10 between the compound-caller bounty and the contract-held treasury share). Direct pays none of this since it never claims.

At current ~1.5% network APR over 90 days: protocol fee ≈ 4,000 × 0.015 × 0.01 × (90/365) ≈ 0.15 ONE. Compounding lift over the same window is a similar order of magnitude. Realized spread should be small and may sit on either side of zero.

At higher APR (5%+) or longer windows (1y+): compounding lift grows roughly quadratically while the fee grows linearly. Vault is expected to pull ahead.

We are not pre-declaring a tight numerical band because the spread depends on a network APR we do not control. What we are committing to: publishing the realized spread daily, declaring the two components above as the only explainers, and publishing any deviation from this decomposition in §5.

4. Conflict-of-interest disclosure

Demo capital source. The 8,000 ONE seeded into the demo wallet was sourced from operator-validator commissions (predominantly Validator.ONE / V1, with a smaller contribution from IPH) which routinely flow into the vault as protocol-side capital.
Symmetric COI on both arms. V1 sits in the mirrored basket at 10% weight on the direct arm AND inside the vault's underlying basket at the same 10% weight. Both arms therefore pay V1 commission proportionally and equally — the COI does not asymmetrically advantage either arm.
Operator stake size. The 8,000 ONE demo is less than 2% of the 489,166 ONE total vault TVL, so the demo's effect on the vault's validator-mix economics is negligible.
No special privileges. The demo wallet uses the same public deposit() and delegate() code paths as any other depositor. There is no admin path, no commission rebate, no reserved buffer slot.

5. Falsification + restart conditions

The experiment is restarted (this epoch retired, a new epoch pre-registered with a new section under §7) if any of:

Vault validator-mix changes by more than 5 percentage points on any validator weight, and the re-mirror SLA is breached. Re-mirroring within 24h keeps the current epoch live; past 24h, retire and restart.
The direct arm claims rewards. Direct is defined as passive. If any collect-rewards or delegate tx ever lands on the demo wallet, the epoch is contaminated. Publish the tx, retire epoch, restart.
A vault upgrade or governance action changes the protocol fee, the validator set, or the compound() interface mid-window. The vault is immutable on these dimensions; restart is the procedure if this assumption ever breaks.
Realized spread is not explained by the §3 decomposition (compounding lift minus 1% fee) within plausible network-APR ranges. We do not stop the experiment — we publish the deviation, the diagnostic, and the corrective action, and run to the end of the window so the full data set is public.

6. Verify it yourself

Demo wallet: 0x599F9f945B6706EEDbe44E8bF40B03BF15Cc85E7
Vault contract: 0x061A7e7e317AcE450940D5aD3372Db0b7C205dE4 — see /properties for the immutable constraint sheet.
Live numbers: the cards on the homepage read on-chain state every 30 seconds. Open browser dev-tools to inspect the raw eth_call and hmyv2_getDelegationsByDelegator responses.
Pre-registration timestamp: the first commit of this page on GitHub is the public anchor proving the parameters above were declared before any funding tx hit the chain.

7. Active epoch

Epoch#1

Pre-registered2026-05-20

Planned start (funding tx)2026-05-20

Planned end2026-08-18

On end-of-window: end-report archived at /transparency/epoch-1-report, a fresh epoch optionally pre-registered as a new sub-section here, and the homepage card resets if a new epoch begins. Past epochs remain navigable from this page.

Methodology questions, methodology gaps, or methodology corrections: open an issue at github.com/stake-one/stake-one. We treat methodology critique as first-class signal.