Proprietary AMMs Lower Solana Swap Costs, Study Finds

A preprint published Sept. 29 finds that proprietary automated market makers, or propAMMs, on Solana delivered substantially lower execution costs for swaps than public AMMs over a one-year sample. The study covers Sept. 1, 2025 through Aug. 31, 2026 and was produced by researchers affiliated with ETH Zurich and Category Labs.

Researchers measured fills on Solana and weighted them by notional value against Bybit’s size-weighted top-of-book USDT microprice, converting via a USDC/USDT midpoint. The analysis includes a quiet-market subset for SOL/USDC trades, defined as cases where the reference price moved less than 1 basis point between five seconds before and one second after a fill. In that subset, propAMMs posted a reference-relative execution-cost proxy of 0.26 basis points, compared with 2.59 basis points for public AMMs.

The paper also reports two-second gross maker markouts of +0.37 basis points for propAMMs and −0.22 basis points for public AMMs. A markout compares the price received on a trade with a later reference price; a positive markout benefits the liquidity provider.

The study distinguishes the experience of a swapper from that of a passive liquidity depositor. A swapper can benefit from tighter quoted spreads and active inventory management. A depositor supplying assets to a public pool faces adverse selection risk when an outside market moves before the pool’s quote, and arbitrageurs can restore price alignment by trading against the pool’s liquidity, creating costs for passive providers.

Industry implementations cited in the paper include propAMMs that adjust pricing and depth based on inventory, quote freshness and incoming flow. One example is a professional operator’s AMM that adapts spreads and available liquidity and applies counterparty controls to manage exposure. Aggregator documentation shows methods to identify retail or lower-risk flow at the routing level, while noting that identifying a trade origin is not the same as independently verifying that every trade is low risk for the maker.

The researchers describe execution anomalies on other chains that illustrate measurement challenges. For a protocol on Base, execution was on average 1.08 basis points worse by trade and 0.56 basis points worse by volume compared with reconstructed previous-block-end quotes, a pattern linked to block timing and fee effects. A routing provider reported that prices sometimes deteriorated between quote selection and settlement because of block timing and spread changes and that it would suspend sources until execution issues were fixed.

Routing design and parity testing are presented as partial remedies for the difference between displayed quotes and executed output. One approach is routers that select executable prices at the time transactions run. Parity checks that compare quoted and executed outputs against the same pool snapshot are required by some integrators; the paper notes that parity testing must match trade size, caller, pool state and applicable charges to be meaningful.

The paper states that lower swap execution costs for traders and the presence of professional market-making practices do not by themselves determine whether passive liquidity providers earn adequate returns. It recommends position-level accounting that includes fees earned, inventory exposure, rebalancing and hedging costs, operating expenses and transaction fees to assess depositor outcomes.