Compare liquidity pools for stablecoins, BTC wrappers and ETH against staked ETH with lending and savings: what each earned on top of simply holding the same coins. Measured on chain over the year to ; newer pools have shorter histories. Past results, not a forecast.
- A narrow range earned 0.6–7.4% a year over holding in the pools with a year of data, full range almost nothing. Where Aave lists both coins, the narrow range beat it in all but two pools. Before gas and the re-set swaps.
- Sky's savings paid 3.6–3.7% a year and beat most dollar pools. Fluid needs no re-set: its LPs earned 0.5–4.2% a year over holding, above Aave in each pool, above Sky only in GHO/USDC.
- Returns varied across windows: bursts of fees lifted the average, and the final month differed from the full study period.
Findings for weekly windows
For the full study period (the last point also the final 30 days) with the position re-set weekly, at the widths the table shows for that; they do not follow the table's settings. Per year.
- Full range earns almost nothing here (0–0.8% a year): the liquidity sits within ±0.1% of the price, so the reference is a narrow range.
- A narrow range re-set weekly earned 0.6–7.4% a year over holding in the pools with a year of data. Where Aave lists both coins, that beat Aave over the same weeks in all but two (Uniswap v4 USDC/USDT 0.001% and 0.0007%, 0.2–0.5 points below): Aave paid 1.6–3.7% for dollars, 0.75–0.8% for ETH with staked ETH, under 0.01% for BTC.
- Sky's savings rate (sUSDS) paid 3.6–3.7% a year on dollars. Of the dollar pools on Ethereum with months of data, six beat it: USDe/USDC, sUSDe/USDT, USDT/USDS, RLUSD/USDC, USDC/USDT 0.01% on V3 and 0.0008% on v4; of the young ones, the dynamic-fee USDC/USDT on v4 and Aerodrome's (USDC/USDG by fees, both staked in AERO). Curve's dollar pools earned 0–0.65%.
- Fluid's pools need no re-set: the protocol keeps one narrow range (0.12–0.4% wide) around the price, and the coins are lent out at the same time. Their LPs earned 0.5–4.2% a year over holding, most of it the lending rate; the fees added 0.04–1.2 points. That beat Aave in each; Sky's savings only in GHO/USDC.
- That is before gas and the swap that re-centres the range each time. The mean is above the median: the fees come in bursts, when a coin moves off its peer. A month differs from the year: over the final 30 days USDC/USDT 0.01% on Uniswap V3 earned 21% a year over holding, the dynamic-fee one on v4 23%, sUSDe/USDT 2.2%.
How to read it
The position. At the start of a window the capital goes into the pool with a range of ±x% around the price of that moment, and is left alone until the window ends. "Re-set every 7 days" is the mean over the windows of 7 days starting each day of the year: as if the position were moved back around the price once a week. Each edge of the range sits on a tick that real positions hold, near ±x%, and must stay held through the window; a pool whose ticks are far apart has few such windows. The fees are read from the pools' own fee counters, so they are exact for such a position.
Range: the width with the highest mean result among ±0.01%, ±0.05%, ±0.1% and ±0.5% that had windows on at least a quarter of the days (10 windows for a pool younger than July 2026). The other widths are in a pool's details.
Income: swap fees and included rewards per the capital, scaled to a year (× 365 / days, simple). Return vs holding: the position at the window end including its income against the same coins simply kept, per year. In these pairs the price barely moves, so it is close to the fees; it falls short when a coin drifts off its peer and the position ends in the cheaper one. Staked ETH drifts up against ETH by its staking, so a pool of it keeps selling it and stays below holding by part of that drift. Median: the middle window. Windows beating holding: the share of the windows above holding. In range: the share of the days, at 00:00 UTC, with the price inside the range.
Without a pool. Aave: what supplying the same coins to Aave v3 (its main market on the pool's chain) earned over exactly the same windows, half in each coin (a Curve pool: in the pool's own mix), from the growth of Aave's income index (getReserveNormalizedIncome) at the same blocks. Like return vs holding, it is what comes on top of holding the coins, so the two columns compare directly. Sky: for a dollar pair on Ethereum, the whole capital in Sky's savings (sUSDS, from its convertToAssets) over the same windows, less the staking the held coins would have earned (sUSDe), so it too is on top of holding. Staking: the staking that the staked coin of a pair earns while simply held (its exchange rate to the underlying: stEthPerToken, getRate, exchangeRate, convertToAssets), half of the capital. It is in both the pool's result and Aave's; it is shown for scale. A coin Aave does not list has no number.
Curve pools are read by their virtual price, which grows by the fees the LPs earn; for the pools with a staked coin (stETH, weETH, sUSDe) less the staking of that coin's share of the pool. Their loss against holding (near zero but for a depeg) is not measured. Windows of 7 days.
Aerodrome pools (from October 2026) have two rows. Unstaked earns the swap fees, as the other pools. Staked earns no fees and AERO instead: here the average staked LP, the AERO the gauge receives each day at its price over the value of the staked liquidity, per year. Nearly all their liquidity is staked.
Fluid pools hold no coins themselves: the coins are lent in Fluid's own lending layer, and LPs hold shares of that supply through Fluid's vaults. The curve is one range around a center price that the protocol moves (for staked ETH, the staking rate), so the position is never re-set. The row is the supply share: its coins per share at the window's ends (getCollateralReserves over the total supply shares) against the same coins kept, which counts both the lending rate and the fees less the loss to the curve; the pool's details split the two. There is no fee counter, so income is not shown separately from the return vs holding. Some pools also have a debt side (LPs who borrow, and whose debt the fees reduce); it is not shown. Fluid's USDC/USDT pools ($32.5M on Ethereum, $11.6M on Arbitrum) have only that side: over the year their LPs paid about 5.9% a year in borrow interest on the debt, and the fees cut it by 0.4 points. TVL is the supply side only.
Dynamic fee: the pool's hook sets the fee of each swap; the fees are read from the counters as in the other pools.
Not included: gas and the swap of each re-set, the risk of a depeg or of a lender, other lenders (Morpho, Spark, Compound), rewards outside the fee counters (FLUID on Fluid's vaults). TVL is GeckoTerminal's on 2026-11-15.