[Insight 05] ·
Notes from the desk
Venue selection as a risk decision
Venue selection in digital-asset markets is usually presented as an optimisation problem. Given a target instrument, the relevant question is: which venue offers the lowest explicit cost, the most observable depth, and the best fill rates? That framing is not wrong. It is incomplete in ways that accumulate quietly.
The fuller framing is that each venue is a bundle of risk exposures, not just a point on a cost curve. The costs that appear on an invoice — maker and taker fees, funding rates, gas — are only the explicit layer. Beneath them sits a set of structural exposures to settlement mechanics, participant composition, and operational concentration that are harder to price but not less real.
Settlement risk is the most structurally distinct. On a centralised venue, a matched trade delivers into a custodied netting cycle: the match is immediate, but the actual transfer of assets depends on the continued operational integrity of an intermediary. On a decentralised venue, settlement is atomic — delivery and payment occur within the same transaction, or neither occurs. That removes custody and counterparty risk from the settlement step but introduces a different set of exposures: public-mempool adversarial dynamics, gas cost uncertainty, and on-chain execution mechanics that interact with price in ways centralised venues do not. A portfolio routing all volume through one settlement architecture is carrying concentration risk that the fee schedule does not capture.
Counterparty composition is the less discussed dimension and often the more consequential one. The adverse-selection cost embedded in a fill depends almost entirely on who is actively quoting on the other side. A venue dominated by professional, flow-informed market makers extracts implicit cost through spread and timing even when the explicit fees are low. A venue with a more retail-skewed or idiosyncratic participant mix tends to carry wider spreads but lower adverse selection per fill. The fee schedule measures the explicit cost of trading at a venue. The participant composition determines the implicit one.
Participant composition is the hardest of the three dimensions to measure. It does not appear in a venue’s documentation, and it shifts over time as maker incentive programs change, new participants arrive, and liquidity conditions evolve. Proxy metrics exist: price impact per unit of size relative to the pre-trade mid; the proportion of fills landing at or through the touch versus closer to mid; and the time-to-adverse-move in the period immediately following a fill. Collecting these across venues in parallel — rather than evaluating each venue against its own historical norms — is what makes the comparison meaningful.
There is a correlated-withdrawal dimension that compounds the picture. Venues that appear independent in normal conditions can narrow liquidity in the same direction under stress: when a large adverse price move creates poor conditions for market makers, liquidity often withdraws from multiple venues simultaneously. Apparent venue diversification can be concentration risk in disguise when the relevant scenario is the one where liquidity is most needed.
The practical implication is that venue selection is a risk parameter with a review cadence, not infrastructure that gets configured once and maintained indefinitely. It belongs in the same conversation as position limits and counterparty guidelines, and it warrants reassessment when participant composition shifts, when settlement mechanics change, or when a venue’s operational risk profile moves materially. Treating it as ops is a reliable way to carry more adverse-selection and counterparty exposure than intended, without it appearing anywhere in the risk reports.