Simulation mode · Paper trading only · Real trading disabled
Research
These are qualitative findings from running the detection and simulation pipeline against live Base and centralized market data. They describe market structure and engineering reality. They are not performance results, and no return figures are published.
Findings
Headline spreads between an on-chain pool and a centralized book appear constantly. Once each side is priced at the size an operator would actually trade, the majority of them are already gone before a single fee is applied.
Price impact on-chain and depth consumption on the book both scale with notional. A dislocation that looks compelling at small size is often uneconomic at meaningful size — and occasionally the reverse, when fixed gas dominates.
Base L2 gas is small in absolute terms but is charged per attempt, not per dollar. It quietly disqualifies the entire small-notional band, which is where naive scanners report most of their hits.
Buying on-chain and selling on a venue is a different trade from the reverse: different fee tiers, different depth profiles, different inventory requirements. Treating a route as bidirectional overstates opportunity.
The interval between detection and execution is when spreads die. Re-quoting before acceptance converts an unknown risk into a measured rejection reason.
Two-leg arbitrage requires capital pre-positioned on both sides. Repeated one-directional flow drains a leg, and the cost of putting it back is a real component of strategy economics.
USDC-quoted stable pairs dislocate in basis points, not percent. They demand tighter cost modeling and larger size to be interesting at all, and they punish fee estimation error hardest.
The most informative artefact the platform produces is not the list of survivors but the distribution of failures: which cost killed which candidate, on which venue, at which size.
Disclosure discipline
Published
Not published
Simulation-only programme · no performance claims are made anywhere on this site
