New Study Reveals Significant Price Discrepancies for Solana Trades Routed Outside Aggregators

A comprehensive academic study published in late September has brought to light significant variations in trading costs across decentralized finance venues, demonstrating that access to the exact same liquidity pool does not inherently guarantee comparable pricing for traders. The research found that realized round-trip spreads for Solana trades routed outside of major aggregators were notably wider at five out of six prominent proprietary liquidity venues, placing the mechanics of transaction routing at the heart of an ongoing debate over price transparency and execution quality in digital asset markets.

The findings indicate that while a liquidity pool may settle trades transparently on a public blockchain, a single operator often maintains absolute control over its quotes. Consequently, evaluating venues by looking solely at pool addresses overlooks a critical component of the transaction lifecycle: the specific pathway through which an order arrives.

The preprint, authored by researchers Ozan Solmaz, Lioba Heimbach, and Jason Milionis and posted on September 29, examined proprietary automated market makers, commonly referred to as propAMMs, across the Solana, Base, and Monad networks. The study’s empirical observations span from September 1, 2025, through August 31, 2026, incorporating later start dates for newer network deployments. Researchers emphasized that the finding on Solana represents a historical association between routing classification and realized spreads, rather than a definitive measurement of what every trading interface delivers in real-time today.

The Same Pool, Different Observed Terms

To understand the dynamics at play, it is essential to distinguish between the various entities involved in decentralized trading. An aggregator functions by searching across multiple liquidity venues simultaneously, possessing the capability to split a single order among them to secure optimal execution; Jupiter serves as a prominent example of such a routing provider on the Solana network. Conversely, a propAMM acts as the underlying venue supplying liquidity, wherein a single operator provides inventory and retains the ability to update prices directly without waiting for an incoming trade to alter the pool’s internal reserves.

The study undertook a detailed comparison of aggregator versus non-aggregator order flows within the six largest measured Solana propAMMs trading the SOL/USDC pair. To reliably estimate a realized round-trip spread, the methodology paired consecutive fills on opposite sides of the same liquidity pool occurring within 25 slots, provided that the trade sizes remained within a factor of two. Furthermore, the criteria mandated that no intervening quote updates could occur from the operator between the fills, and both legs of the transaction had to belong to the same routing class.

By comparing these routing classes within analytical cells strictly defined by specific pools, days, and trade-size buckets, the researchers ensured a far more informative test than simply comparing unrelated venues or orders of disparate sizes. Although the methodology pairs observed historical trades rather than tracking an individual customer’s actual buy-and-sell journey, the controls successfully standardized key variables.

Across the venues analyzed, the data revealed a stark divergence in execution quality. At HumidiFi, aggregator median spreads stood at 0.09 basis points compared to 0.86 basis points for non-aggregators, yielding a matched non-aggregator premium of 0.4 basis points, with 76% of matched cells showing wider non-aggregator spreads. BisonFi exhibited aggregator medians of 0.71 basis points and non-aggregator medians of 1.01 basis points, featuring a matched premium of 0.1 basis points and 76% wider non-aggregator cells. Tessera recorded aggregator medians of 0.9 basis points against 0.93 basis points for non-aggregators, with a 0.1 basis point matched premium across 68% of matched cells.

The disparities widened substantially at other venues. SolFi V2 posted aggregator median spreads of 3.88 basis points and non-aggregator medians of 4.9 basis points, resulting in a matched premium of 1.58 basis points and 82% of cells showing wider non-aggregator spreads. GoonFi demonstrated the most pronounced divergence among the primary group, with aggregator medians at 0.78 basis points and non-aggregator medians at 4.39 basis points, translating to a substantial matched premium of 3.59 basis points and 83% of cells registering wider non-aggregator spreads. A single exception emerged in the dataset: ZeroFi recorded aggregator medians of 1.52 basis points and non-aggregator medians of 1.5 basis points, producing a slightly negative matched premium of -0.05 basis points and only 41% of cells showing wider non-aggregator spreads.

In this context, one basis point equates to 0.01 percentage point. The distinction between overall medians and matched premiums is vital for accurate interpretation. For instance, Tessera’s overall medians were 0.9 and 0.93 basis points, whereas its median matched premium was calculated at 0.1 basis points, a figure derived by comparing the two routing classes internally within each specific cell prior to summarizing those differences.

For the first five venues in the study, non-aggregator spreads were consistently wider in the vast majority of matched cells, with reported exact sign-test p-values falling well below 0.001, underscoring strong statistical significance. Researchers noted that these figures represent round-trip price gaps rather than individual swap fees, nor do they formally identify ordinary retail customer behavior, as non-aggregator flow can inherently possess different characteristics than routed aggregator flow. While matching across pools, days, and sizes successfully mitigates some structural differences between the groups, it does not constitute a randomized experimental test of execution changes resulting purely from switching interfaces.

The underlying economic rationale for these disparities largely stems from the management of stale prices. A passive liquidity pool is perpetually exposed to the risk of losing capital when an external trader acts on a broader market movement before the pool’s internal pricing mechanism has caught up. PropAMM operators mitigate this exposure by issuing price updates independently of executed swaps, thereby reducing their risk profile whenever these updates successfully outpace incoming trade requests.

Solana study finds wider spreads outside aggregators at five of six sampled pools

This operational capability ultimately benefits liquidity operators by minimizing losses incurred from stale pricing. Within the examined Solana sample, aggregator flow demonstrated a strong historical association with tighter matched spreads at the vast majority of the measured venues.

To provide deeper technical context, the research paper also highlighted a mechanism example observed on the Base network. In an implementation of Tessera decoded by the study, researchers identified six distinct fee components governing transactions: base, age, priority, mode, default, and address. The identity of the transaction sender, alongside the specific smart contract invoking Tessera’s interface, directly influences how default and address parameters are applied, creating potential exemptions and penalties.

This finding illustrates how a proprietary venue can incorporate caller identity directly into execution pricing. However, researchers cautioned that this observation serves as evidence specific to the inspected Base implementation rather than definitive proof that Tessera’s Solana deployment—or any other Solana propAMM—utilizes identical code or fee structures.

A related concern involves the persistence of attractive quotes through to final settlement. The researchers documented instances on Base where Tessera fees decreased near the end of blocks and increased near the beginning of blocks, helping to explain observable discrepancies between a previous block’s quoted price and the ultimate fill. However, the study explicitly lacked historical Solana quote data that would be necessary to corroborate an analogous quote-to-fill gap on the Solana network.

These findings align broadly with prior industry observations. In March, aggregator 0x published an investigation detailing Base quote-to-fill discrepancies associated with Flashblock timing and spread adjustments, though those findings remained specific to that network and originated from an interested routing provider.

Wallet Fees Add Another Layer of Complexity

Beyond underlying pool spreads and routing mechanics, user-facing wallet charges introduce an additional layer of cost that directly impacts the final amount received by a trader. Even when an underlying liquidity pool offers highly competitive terms, a wallet application’s internal fee structure can significantly alter the net financial outcome of a transaction.

Current disclosures from major wallet providers illustrate this dynamic. Documentation from Phantom outlines a standard fee applied to the majority of swaps alongside variable network fees, with specific exceptions designated for certain account flows and perpetual balance deposits. Such wallet-level fees represent an entirely different measurement category than the round-trip spreads examined in academic literature.

In decentralized finance terminology, price impact describes the degree to which an order’s own size affects its execution price, whereas slippage refers to the price movement occurring between the initial quote and final execution. Setting a slippage tolerance establishes a boundary for how much price deterioration a transaction is permitted to endure; raising this threshold does not impose a fixed fee, but it creates the possibility of accepting a worse fill.

API documentation from Jupiter further demonstrates why interface labels alone can leave critical questions unanswered regarding transaction costs. Jupiter’s managed order path enables multiple providers to compete, whereas its customizable build path utilizes specific request-for-quote liquidity mechanisms distinct from on-chain propAMM pools. While Jupiter documents specific platform fees for certain swap categories on its managed path alongside separate integrator and network costs, these disclosures describe available integrations rather than identifying the precise route utilized by any given wallet swap.

Consequently, attempting to combine a historical matched spread premium with a current wallet fee can produce misleading financial estimates. Such a calculation would improperly mix a paired round-trip market statistic with a retail charge on an isolated swap, without establishing whether specific costs are truly incremental or already absorbed within the delivered token amount.

Major routing providers continue to pursue technological solutions to narrow the gap between quoted expectations and completed settlements. Jupiter has outlined execution-estimation frameworks that incorporate observed historical execution data and simulation, alongside architecture that allows candidate liquidity pools to be re-quoted on-chain at the moment of transaction execution, enabling the system to dynamically adjust allocations if a better output becomes available. Ultimately, industry experts suggest that the most reliable metric for evaluating trading efficiency remains the completed trade in its entirety: the specific route and venue utilized, all applicable fees, and the precise amount successfully delivered under comparable market conditions.