Curve: decentralised exchanges rating breakdown
Stablecoins and pegged assets, 31 chains
Still the venue where large stablecoin and pegged-asset swaps belong: its invariant is built for assets that should trade near parity, and for that specific job nothing here prices better. At 0.1040% realised across all pools it is mid-priced, and its $1.56bn of TVL across 31 chains is the widest genuine deployment after Uniswap.
Still the venue where large stablecoin and pegged-asset swaps belong: its invariant is built for assets that should trade near parity, and for that specific job nothing here prices better. At 0.1040% realised across all pools it is mid-priced, and its $1.56bn of TVL across 31 chains is the widest genuine deployment after Uniswap.
How the score is built
Each criterion is scored 0–10 and weighted. The median column is the middle score across every entry in this ranking, so a row reads as a position rather than a number.
| Criterion | What we measured | Weight | Score | Category median | Rank | Weighted gap |
|---|---|---|---|---|---|---|
| Has it been drained, and did anyone get paid back? | — | 30% | 4 | 5 | 6 of 9 | -0.30 |
| What traders actually paid | 0.104 what traders actually paid · computed · 2026-08-27 · source | 25% | 7 | 7 | 5 of 9 | 0.00 |
| Depth and volume | $3.05bn 30-day volume · sourced · 2026-08-27 · source$1.56bn total value locked · sourced · 2026-08-27 · source | 25% | 8 | 7 | 3 of 9 | +0.25 |
| Chains and assets | 31 chains deployed · sourced · 2026-08-27 · source | 12% | 9 | 6 | 3 of 9 | +0.36 |
| How hard the capital works | 2.0× per 30 days capital turnover · computed · 2026-08-27 | 8% | 3 | 7 | 8 of 9 | -0.32 |
Measured 27 August 2026 · weights and method · decided by chains and assets, worth +0.36 points against the median
Has it been drained, and did anyone get paid back?: 4/10
Contract exploits with dates, amounts and — the question every other comparison skips — whether users were actually made whole. A DEX is a set of contracts holding pooled money, so this is the first question, not the fifth. Audit count is recorded but never scored on its own: Balancer V2 was among the most audited contract systems in DeFi and lost $128m to a rounding error in November 2025.
Scored 4 of 10 against a category median of 5, which places it 6th of 9 among decentralised exchanges on this criterion. At a 30% weight that is 0.30 points below the median contribution of the weighted total. The best score in the category is 9, the worst 1.
What traders actually paid: 7/10
Computed by us, not read off a fee-tier page: thirty days of protocol fees divided by thirty days of volume, per venue, from public data. That single division exposes what routing, tier mix and pool design really cost, and it disagrees with the advertised tiers everywhere. It is not the whole cost of a trade — slippage and gas sit on top — but it is the part every comparison misstates.
Scored 7 of 10 against a category median of 7, which places it 5th of 9 among decentralised exchanges on this criterion. At a 25% weight that is exactly level with the median of the weighted total. The best score in the category is 10, the worst 1.
Depth and volume: 8/10
Total value locked and thirty-day volume, both from public data at a stated timestamp. Depth is what decides slippage on a real trade, and volume is what proves the depth is usable rather than parked.
Scored 8 of 10 against a category median of 7, which places it 3rd of 9 among decentralised exchanges on this criterion. At a 25% weight that is 0.25 points above the median contribution of the weighted total. The best score in the category is 10, the worst 2.
Chains and assets: 9/10
The number of chains where the venue is actually deployed, weighted by whether volume genuinely flows there. A deployment on forty chains that trades on two is coverage on paper.
Scored 9 of 10 against a category median of 6, which places it 3rd of 9 among decentralised exchanges on this criterion. At a 12% weight that is 0.36 points above the median contribution of the weighted total. The best score in the category is 10, the worst 2.
How hard the capital works: 3/10
Thirty-day volume divided by total value locked — how many times each dollar of liquidity turned over. Nobody publishes this, and it separates a venue whose pools are genuinely used from one sitting on idle incentive-farmed capital. It also predicts fee income per dollar of risk taken by liquidity providers.
Scored 3 of 10 against a category median of 7, which places it 8th of 9 among decentralised exchanges on this criterion. At a 8% weight that is 0.32 points below the median contribution of the weighted total. The best score in the category is 10, the worst 3.
Its nearest neighbours in this ranking
Incidents priced into this score
- 2023-07-30 — A reentrancy bug in specific Vyper compiler versions (0.2.15–0.3.0) drained four Curve pools. Roughly $73.5m was taken; about 73% was returned by 7 August 2023 after white-hat negotiation and a 10% bounty offer. The flaw was in the compiler, not in Curve’s pool logic. [users made whole: partial] [source]
Questions about this score
What caused the Curve hack in 2023?
+
A reentrancy bug in specific versions of the Vyper compiler — 0.2.15 through 0.3.0 — rather than a flaw in Curve's own pool logic. Roughly $73.5m was drained from four pools on 30 July 2023, and about 73% came back within a week after negotiation and a 10% bounty offer. The location of the bug matters: an audit of the protocol's source would not have caught a fault in the tool that compiled it.
Is Curve still the best place to swap stablecoins?
+
For large size in pegged assets, yes — that is what its invariant is built for, and nothing else in this rating prices those swaps better. At 0.1040% realised across all pools it sits mid-table on cost, and its capital turnover of 2.0× a month is the lowest of the major venues, which tells you a great deal of its TVL is parked for incentive reasons rather than being traded.
Who controls Curve?
+
Long-locked token holders, through vote-escrow governance. Pool contracts are not upgradeable by an admin and parameter changes run through the DAO with delays, which is a stronger position than most of this category — but the voting power is concentrated among holders who locked early and long, so 'the DAO decides' means something narrower here than it sounds.