Non-custodial swap aggregators show users the best single-exchange rate, but is that actually optimal execution? This talk presents empirical findings from a live BTC→XMR aggregator across 23+ exchanges, examining whether splitting trades can improve outcomes.
The answer: yes, but only sometimes. Across 2,714 logged events, split routing gain scales with trade size, averaging 0.38% at 8 BTC versus 0.10% at 0.5 BTC, with a >1% tail emerging at 2 BTC and above. Zero events produced a negative outcome.
The underlying mechanism is the liquidity cliff: every exchange has a capacity ceiling beyond which their rate degrades. Traditional aggregators query at the full amount and never see it. Exploiting it requires continuously pre-sampled sub-amount rate curves, infrastructure most aggregators don't have.
We present the detection methodology, the gain distribution, and when split routing is worth surfacing and when it isn't.