|
ClearTrace
DEX Execution Insights
August 20, 2026 · Ethereum mainnet
|
|
Lead finding
A router's headline revert rate is mostly bots. A real user hits far fewer.
A revert rate counts the transactions that failed. On a busy router most of those failures are not people: they are searchers firing atomic arbitrage that reverts by design the moment the profit is not there. Count only the senders that behave like genuine users and the number drops sharply. Over the past 7 days on Ethereum, the user-submitted revert rate sits under 1% on every major router we track, and the headline overstates it by as much as 6x. Below, the headline rate against the user-only rate, over all trades on each router.
| Router |
Headline revert |
User-only revert |
| OKX | 5.35% | 0.85% |
| OpenOcean | 4.14% | 0.74% |
| Sushi | 2.01% | 0.34% |
| Uniswap (venue default) | 1.96% | 0.56% |
| KyberSwap | 1.37% | 0.53% |
| ParaSwap | 1.27% | 0.29% |
| 1inch | 1.07% | 0.28% |
On-chain over all Ethereum trades, 7-day window ending 2026-08-19. Headline is every failed transaction the router sent; user-only strips senders classified as bots or searchers (methodology v6). OKX is the widest split: a 5.35% headline is 0.85% once the arbitrage spam is set aside, a 6x cut. These are router venues, where the user submits the swap and a failed swap is a transaction we can count.
The takeaway is not that one router is safer than another. It is that the headline number is the wrong number to read. Ranked on what real takers actually experience, the spread across every major router is narrow, and all of it is under 1%. The reliability gap that traders worry about is largely an artifact of who else is in the transaction set.
|
|
|
Also this period
|
The lowest user-only revert rates were 1inch at 0.28% and ParaSwap at 0.29%. Sushi came in at 0.34%. Uniswap's direct default sits at 0.56%. The full router set ran from 0.28% to 0.85%.
|
|
You cannot rank every venue this way. CoW settles in a batch auction (0.12% headline reverts) and 1inch's Fusion flow fills as an intent, so an order that never fills never becomes an on-chain transaction. Their near-zero revert is a structural property of how the trade clears, not a reliability score. Ranking it against a router would be a category error.
|
|
The quote-accuracy sample is now 323 rated cells. Each is built from at least 30 fork-simulated fills over a 7-day span; the sample now runs 58.6 days deep, from 66,780 realized simulations. On identical pairs and sizes, most rated routers hold their quotes tightly. OpenOcean is the widest at a median 10.2 bps, but that gap is its published fee (1 bp on stables, 10 bps on volatile pairs, paid to a named collector address), not execution shortfall. Strip the fee and its organic slippage is 1.3 bps, in line with the rest. ParaSwap charges a flat 1 bp the same way. The gap is what a taker actually receives against the quote, so the fee belongs in it; it is simply a different thing from missing the quote.
|
|
Coverage on Ethereum holds at 95.8% of tracked DEX volume attributed to a named frontend or router. This is the cleanest of the four chains we track: it carries no single-transaction pricing outliers this period.
|
|
|
|
The full aggregator leaderboard, the per-router revert breakdown and the quote-accuracy sample are on the ClearTrace dashboard. The methodology notes explain how the user-only revert rate, the execution-model classification and the rated-cell bar are computed.
|
|
ClearTrace · Cross-chain DEX attribution · Aggregator-leaderboard and quote-sampler data, Ethereum, 7-day on-chain window ending 2026-08-19. The headline revert rate is every reverted transaction a router sent over the window. The user-only revert rate (methodology v6) is the same measure over just the senders classified as genuine users, with bot and searcher senders removed. Revert rates are comparable only across router venues, where the user submits the swap; a batch auction (CoW) or an intent venue (1inch Fusion) clears the trade off-chain, so an unfilled order never reverts on-chain and its rate is not a reliability signal. The quote-accuracy gap is a venue's quoted output against the output a fork simulation of its returned route produces; it excludes gas, mempool competition, inclusion delay and real MEV, and cannot reproduce off-chain RFQ legs, so it is not a complete all-in trading cost. Where a venue charges an explicit protocol fee, that fee lands inside this gap, so a wide gap is not by itself evidence of an inaccurate quote. Rated means at least 30 realized samples and a 7-day span. Every figure above is reproducible from ClearTrace's published data; the live leaderboard moves with each sync.
|
|