|
ClearTrace
DEX Execution Insights
July 16, 2026 · Ethereum mainnet
|
|
Correction · added July 30, 2026
The attribution-coverage figures in this issue were overstated. Contracts our pipeline had explicitly failed to identify, recorded internally as "Manual Review Required", were counted as attributed volume rather than as unattributed. Recomputed from the same data this issue was built on, cross-chain coverage was 89.3%, not the 93.1% stated below.
Per chain, corrected: Ethereum 89.9% (not 94.4%), Base 88.3% (not 90.5%), Arbitrum 85.8% (not 88.4%), Optimism 82.8% (not 83.8%).
The text below is unchanged from what was sent. Every issue from June 28 to July 23, 2026 was affected, by roughly 3 to 4 points. The direction we reported holds: coverage did rise over that period, from 87.8% to 90.6% on the corrected basis, and every chain remained above our 85% internal bar. The cause is fixed and the figure now comes from a single shared definition. Our full corrections log is available on request.
|
|
|
Lead finding
An honest quote is not a cheap trade
ClearTrace measures two separate things per aggregator: whether the price you are quoted is the price you get (quote accuracy, from fork-simulated samples), and what that trade actually costs against mid (execution cost, from every on-chain fill). This period the two barely correlate. Six of the seven rated aggregators quote honestly, with a median gap between quoted and realized price under 2 bps. Yet the realized cost of those same trades ranges from 4.36 bps (1inch) to 23.96 bps (Sushi), a spread of roughly 20 bps on overlapping major pairs.
The Uniswap direct-DEX baseline, the cost of just swapping straight through Uniswap, sat at 10.04 bps. Only 1inch (4.36), Bebop (8.80) and Odos (8.86) came in under it. A tight quote tells you the router is not overselling its price; it does not tell you the route it picked was the cheapest one. Both numbers are worth checking, and they answer different questions.
|
|
|
This period in numbers
| |
Quotes hold up, with one exception
Across seven rated aggregators, the median gap between quoted and realized price is 0.0 bps for 1inch, Bebop and Sushi, 0.03 bps for Odos, 1.0 bps for Paraswap and 1.81 bps for KyberSwap. OpenOcean is the widest at 11.48 bps, which is its published fee (1 bp on stables, 10 bps on volatile pairs, paid to a named collector address) rather than execution shortfall. Source: ClearTrace quote sampler (fork-simulated, rated cells only), generated 2026-07-16.
|
| |
Execution cost spans 4 to 24 bps
Median on-chain execution cost over all trades: 1inch 4.36 bps, Bebop 8.80, Odos 8.86, KyberSwap 10.20, OpenOcean 14.73, Paraswap 17.64, Sushi 23.96. These are observed medians and partly reflect each router's trade mix, but only three beat the 10.04 bps Uniswap baseline. Source: on-chain slippage (onchain_slippage_bps), synced 2026-07-15.
|
| |
Revert rates only compare within a model
Among order routers, the revert rate over genuine users (methodology v6) runs from 0.26% (1inch) to 2.52% (Odos). CoW Protocol's near-zero on-chain revert is structural, not a reliability edge: a solver submits the trade, so an unfilled order never becomes a reverted user transaction. Do not rank revert across execution models. Source: onchain_user_revert_rate_pct, 2026-07-15.
|
| |
214 rated cells, 93.1% attribution coverage
The sampler has now taken 18,672 fork-simulated quote samples over 23.4 days (91% realized), of which 214 cells have cleared the rated bar (at least 30 realized samples and a 7-day span). Cross-chain attribution coverage sits at 93.1% (Ethereum 94.4%, Base 90.5%, Arbitrum 88.4%, Optimism 83.8%). Source: quote sampler report + attribution coverage, synced 2026-07-15.
|
|
|
|
Explore the full aggregator leaderboard, chain-by-chain coverage, and live attribution data at the ClearTrace dashboard. The methodology notes explain how execution cost, quote gaps, revert rates, and rated-cell thresholds are computed.
|
|
ClearTrace · Cross-chain DEX attribution · Data as of 2026-07-16 (quote sampler) / 2026-07-15 (on-chain attribution).
Execution cost and revert rate are on-chain medians over all trades; the Uniswap baseline is the direct-DEX median that aggregators are compared against. Quote-gap figures are rated (at least 30 fork-simulated realized samples and a 7-day span). Revert rates are not comparable across execution models.
Figures are sourced from on-chain trace data and the ClearTrace live quote sampler. No protocol claims are fabricated.
|
|
Correction · added October 1, 2026
The third bullet above describes the user-only revert rate (methodology v6) as “the revert rate over genuine users”. That is not what the filter does. It performs no classification of who a sender is. It keeps the senders whose own transactions to these routers succeeded at least 90% of the time over the window and drops the rest, a threshold on each sender’s own outcomes. So the figure is bounded under 10% by construction, and a sender whose few transactions included one failure is removed from the numerator and the denominator alike.
Read every user-only figure in this issue as the failure rate among consistently-succeeding senders, not as the rate people experience and not as the share of all attempts that fail. The headline figure beside it is every reverted routing transaction after the separate v5 bot filter. The senders the threshold removes carry a disproportionate share of the reverts (on the 2026-09-16 data, about 5% of transactions and roughly half of all reverts); some of them are searchers, but the filter does not identify them, so the gap between the two columns is not a measurement of bot traffic. No figure in this issue changes.
The text above is unchanged from what was sent. The filter is set out in full on the methodology page. Our full corrections log is available on request.
|
|
|