Skip to main content
September 12, 2026 · Robinhood Chain Route led in 130 of 151 valid Kyber comparisons across five pairs, both directions, four sizes and four rounds.
These were unsigned simulations using experimental contracts, not completed trades or production results. Comparisons exclude Route’s fee and the gas used to collect it; other execution gas and pool/provider charges remain included.

​
Results

Average Route output edge: plus 11.24 basis points against Kyber across 151 valid comparisons; plus 2.96 basis points against Jumper across 12.
1 bp = 0.01%. Median edges were +2.42 bps against Kyber and +0.63 bps against Jumper. Ties fall within ±0.1 bps. Averages include losses and give each valid comparison equal weight. Route’s results favored it in all four Kyber rounds. ROUTE → ETH was its weakest direction, with seven wins and seven losses. Jumper’s 12 valid comparisons are too few for a broad conclusion. Average Route edge against Kyber by direction. ROUTE to PONS leads at 57.35 basis points; ROUTE to ETH is smallest at 0.86 basis points. Exact values follow in the expandable table.
Each direction had 16 planned cases. Missing or failed comparisons are excluded, never counted as wins.

​
Route vs Kyber

​
Route vs Jumper

​
Speed

Recorded Route quote latency was 1.98 seconds median and 3.49 seconds P95. On matched Kyber cases, Route’s median was 1.783 seconds versus 0.857 seconds for Kyber’s build plus simulation. This study does not show a Route speed advantage.
The study reports all 160 Route execution simulations and fee-gas controls passed. Comparisons used the same parent state and checked exact input, recipient balance changes, minimum output and gas.The edge is the difference in output after included gas, divided by the competitor’s output before gas, multiplied by 10,000. Fee-collection gas was measured by replaying the same transaction with its fee enabled and disabled; the median deduction was 85,263 gas. Latency measurements retain the original fee-enabled execution checks.Kyber had five rate limits and four reverts. Jumper had two reverts, one rate limit and 145 skipped cases during cooldown. Meta remained unmeasured in this study.Synthetic balances, approvals and fresh prototype receiver storage were used. Token-output gas conversion is advisory. Separate approvals, rollup charges and browser latency are excluded. Repeated samples cover one market window; they do not establish universal superiority. Broader competitor coverage, independent time windows and production validation remain open.
Study post · Earlier MetaMask study · Earlier Uniswap study Figures are reproduced from the supplied study summary. Earlier studies use different methods and remain separate.