Case 7 · On-chain yield
Case 7 — On-chain yield: the APY you see vs the APY you get
Question. A vault advertises 8.44% net APY. What did a holder actually earn over the last 30 and 90 days?
Method. For an ERC4626 vault, convertToAssets(10**shareDecimals) returns
assets per share. Sample it at the latest block and at blocks ~7/30/90/180
days back; read each block's timestamp; annualize the share-price ratio:
(pps_now / pps_t) ** (365.25 * 86400 / (t_now - t)) - 1
No indexer, no vendor APY history, no trust in the issuer's dashboard.
Two public vaults, same method.
A. sUSDS (Sky Savings USDS) — an honest headline
| Item | Value |
|---|---|
| Vault | 0xa3931d71877C0E7a3148CB7Eb4463524FEc27fbD (Ethereum) |
| totalAssets | 4,866,313,574 USDS (~$4.87B) |
| price per share | 1.11345598 |
| headline APY | 3.80% (DeFiLlama, 2026-10-20) |
| Window | pps | Realized APY |
|---|---|---|
| 7 days | 1.11265706 | 3.8027% |
| 30 days | 1.11013524 | 3.6871% |
| 90 days | 1.10377501 | 3.5935% |
| 180 days | 1.09410721 | 3.6074% |
Verdict. The number is honest: 7-day realized equals the headline. But the headline is a point-in-time rate. Over 30/90 days a holder earned 0.11/0.21 pp less. Using 3.80% as a three-month expectation overstates it.
B. Yearn OG USDC (ymvOG-USDC) — a structural gap
| Item | Value |
|---|---|
| Vault | 0xF9bdDd4A9b3A45f980e11fDDE96e16364dDBEc49 (Ethereum) |
| totalAssets | 10,240,839 USDC (~$10.24M) |
| price per share | 1.079601 USDC |
| headline netApy | 8.4368% |
| netApy excluding rewards | 7.9207% |
| vault's own avgNetApy | 6.1445% |
| performance fee | 5% |
| Window | pps (USDC/share) | Realized APY |
|---|---|---|
| 7 days | 1.078423 | 5.8413% |
| 30 days | 1.073987 | 6.5244% |
| 90 days | 1.064046 | 6.0429% |
| 180 days | 1.051274 | 5.5222% |
Verdict. The headline overstates the 30-day realized return by 1.92 pp and the 90-day by 2.39 pp. Three separate sources:
- Reward tokens (0.52 pp).
netApycapitalizes token rewards, but the share price only accrues base yield. A holder who does not claim and sell the rewards receivesnetApyExcludingRewards, notnetApy. - Point-in-time vs period. Even excluding rewards, 7.92% is 1.40 pp above
the 30-day realized 6.52%. The vault's own
avgNetApy(6.14%) is the comparator that actually matches the realized series. - Gross vs net of fees. The 5% performance fee is already inside the
share price; the gross
apy(8.35%) is not.
Why this matters
Allocators compare headline APY. The realized share-price series is what a holder actually received. The gap is not one number but three measurable sources — and none of them requires trusting the issuer's dashboard.
What this does not prove
- One vault per case, one source for each headline, four lookback windows.
- The vault contracts, reward-claim path, and fee-accrual mechanics were not audited.
- Scope is the two vaults and the dates shown; the numbers do not generalize.
- The public archive endpoint has no SLA; re-running later will move the latest-block window.
Open lead (not a finding)
DeFiLlama reports SENPYUSDMAIN at $388M TVL / 5.34% (2.33% base + 3.00%
reward). Morpho's own API lists the Sentora PYUSD vaults at near-zero TVL and
3.4–3.8% net. Two public sources disagree by ~$388M on the same object.
Resolving it needs the vault address and on-chain totalAssets; until then it
is a lead, not a conclusion.
Reproduce
python3 cases/repro_case7_yield.py --vault 0xa3931d71877C0E7a3148CB7Eb4463524FEc27fbD --reported 3.80
python3 cases/repro_case7_yield.py --vault 0xF9bdDd4A9b3A45f980e11fDDE96e16364dDBEc49 --reported 8.4368