Why we’re posting. Proposals that put DAO money into liquidity pools often quote the pool’s fee rate (APR). We want to show, with real numbers, why that rate on its own can’t tell you whether the pool made money, and to suggest three simple questions anyone can ask when they see one. This isn’t about any particular proposal, and nobody paid for it.
The short version. A pool’s fee rate only counts the money coming in. It leaves out what you lose when prices move. In our data, that loss took back all of the fees on Ethereum-style chains, and about half of them on Solana.
What the fee rate leaves out
When you put two tokens into a pool, you earn trading fees. That part is real, and it’s what the fee rate shows. But when one token’s price falls, the pool automatically ends up holding more of that token. So compared with simply keeping your two tokens in a wallet, you lose some money. The fee rate never shows this loss. It can’t go below zero.
So we measured both parts together on real pool history: fees earned, minus the loss from price moves, compared with just holding the two tokens.
1. Same period, two very different results.
Both rows cover January 2025 to July 2026. Each pool counts equally.
| pools | fees | loss from price moves | what was left | |
|---|---|---|---|---|
| Ethereum-style chains (Ethereum, BSC, Base, Polygon, Arbitrum) | 59 | ~6.4% a year | ~8.2% a year | about −1.8% a year |
| Solana | 56 | ~29.9% a year | ~15.1% a year | about +14.8% a year |
On the Ethereum-style chains, price moves took back all of the fees and a bit more. Only 24 of the 59 pools did better than just holding the tokens. On Solana the pools had far more trading, so fees were much higher: price moves took back about half, and 42 of 56 pools did better than holding. Looking at the fee rate alone, you couldn’t tell these two situations apart.
2. Pools of smaller tokens pay more fees, but not enough to cover the bad cases.
We compared 102 pools of smaller tokens with 18 pools of major tokens on Solana, over the same recent three months. The typical pool did about the same in both groups. The bad cases did not: one smaller-token pool in ten lost 7.5% or more compared with holding, and the worst lost 59%. Among the major tokens, the worst lost 5%. The extra fees were only about 0.4% over the three months. Averaged over all pools, the smaller-token pools lost money (−0.6%) while the major-token pools made money (+1.5%). The real picture is probably worse, because pools that shut down early were already missing from the data.
3. The same thing, live. Since 20 August 2026 we’ve been running a paper test: $1,000 in each of 19 Solana pools, no real money. After 48 days the pools have earned +3.7% in fees, and price moves have taken back −3.5%. After the cost of getting in, the result is about −0.1%. Price moves took 94% of the fees. Seven weeks is far too short to judge the pools. But it shows how far apart “fees earned” and “money made” can be.
Others are finding the same thing. An independent open-source project, REALIZED, reports that on Uniswap v3 “34–62% of pools advertised a positive APR while LPs went backwards”.
Three questions to ask when a proposal quotes a fee rate
- Compared with what? Does the number subtract the loss from price moves, compared with just holding the two tokens?
- Which pools? A list of pools that exist today leaves out the ones that shut down. Were all the pools in the number already there at the start?
- Whose data? Pool data from data sites contains errors. One site listed a pool’s fee rate as 172,635% for a single day, when its normal rate was under 1%. One bad day like that can make a whole average look great. It happened to us: it inflated one of our own early results until we caught it.
Limits, plainly. Only one Arbitrum pool is in the first row, so we can’t give Arbitrum its own number. Going back to 2022 with 108 Ethereum-style pools, the result is about +1.4% a year: still close to nothing. How you average matters too: if long-lived pools count for more, the first row comes out near zero (+0.4%) and Solana higher (+17%), but the picture doesn’t change. We only measured standard pools. Many Arbitrum pools are “concentrated” pools, where your money sits in a chosen price range. Those lose more from the same price moves and need a different calculation, so we left them out rather than guess. We only included pools whose tokens have a reliable price history (about one in ten). Numbers are before gas and transaction costs. These are past results, not a forecast.
Who we are: Arcobaleno Research. We check performance claims (backtests and track records) for a fixed fee, and we publish our own mistakes alongside our method. Happy to answer questions about the method here.