TL;DR
This paper introduces a rigorous framework to measure "Price Improvement" (PI) in onchain Order Flow Auctions (OFAs) like UniswapX and 1inch Fusion. By internalizing gas costs and using counterfactual simulations, the authors find that these auction-based systems provide a statistically significant boost to traders—averaging 4–5 basis points—primarily driven by superior routing and access to deeper liquidity.
What Problem Does This Paper Address?
While Automated Market Makers (AMMs) have revolutionized decentralized finance, they suffer from "hidden" costs: fragmented liquidity, high gas fees, and value extraction by adversarial actors (MEV). Order Flow Auctions (OFAs) were designed to solve this by outsourcing trade optimization to specialized "solvers" or "fillers" who compete to give users the best price.
However, proving that OFAs actually provide better deals is difficult. Standard price comparisons often ignore gas fees, which can account for over 90% of the cost for small trades. Furthermore, there hasn't been a consistent way to determine what a "fair" baseline price would have been if the user hadn't used the auction. This paper fills that gap with a formal mathematical definition of price and an attribution model to explain why one platform might outperform another.
Methods
The authors propose three core innovations:
- Gas Internalization: They redefine "price" to include the transaction fee. This allows for a "head-to-head" comparison between "gas-free" transactions (where solvers pay the gas) and traditional swaps (where users pay the gas).
- Counterfactual Baselines: To see if a trade was "good," they simulate what would have happened if that same trade had been executed via a standard AMM (Uniswap Classic) at the exact same moment. They use the Uniswap Routing API and the Tenderly Simulator to generate these baselines.
- PI Attribution Model: Using a Taylor Series expansion, they decompose Price Improvement into three controllable factors:
- Routing Efficiency: Accessing better liquidity pools.
- Gas Optimization: Writing more efficient smart contract code.
- Priority Fee Settings: How much is paid to miners/validators for inclusion speed.
Datasets
The study analyzes historical onchain data from November and December 2023 on the Ethereum mainnet.
- Scope: Specifically focused on WETH-USDC trades, which represent a massive portion of DEX volume.
- Sources: Data was pulled via Dune Analytics, covering four settlement paths: 1inch Aggregator, 1inch Fusion, Uniswap Classic, and UniswapX.
- Sample Size: Over 15,000 swaps across the four platforms.
Key Insights and Ideas
- The "Routing" Advantage: For large trades, the vast majority of price improvement comes from better routing. Solvers in OFAs can tap into private liquidity or complex paths that standard AMM routers might miss.
- The Gas/Liquidity Trade-off: Adding more liquidity sources usually increases gas costs. The paper shows that for small trades, the extra gas cost of a complex route often outweighs the price benefit, which is why OFAs sometimes struggle to beat simple AMM swaps for "retail-sized" orders.
- Systematic Bias in Simulations: The authors discovered that simple simulations often underestimate gas usage. they introduced a regression-based correction ($\beta_1$ parameter) to ensure their baseline comparisons were statistically sound.
- Non-linear Gains: The "corrections" term in their Taylor expansion (representing non-linear interactions between gas and fees) was significant for UniswapX and 1inch Fusion, suggesting that the way these protocols internalize costs is more complex than a simple linear sum.
Key Results
- Average Improvement: Auction-enhanced interfaces (UniswapX and 1inch Fusion) provided an average price improvement of 4–5 basis points (bps) compared to the baseline.
- Statistical Significance: While small trades ($<$ $25k) showed high variability, the price improvement became consistently positive and stable as trade size increased.
- Platform Comparison: At a 0-block offset (execution time), Uniswap users saw ~4.6 bps improvement, while 1inch users saw ~4.3 bps. Both were significantly better than the "Classic" non-auction counterparts.
- Liquidity vs. Gas: For large trades, increased liquidity was the primary driver of PI. For small trades, gas efficiency and pool selection were the dominant factors.
Limitations and Future Work
- Simulation Timing: The authors currently simulate at the "end of the block." This doesn't perfectly capture intra-block dynamics or "front-running" that might occur in the real world.
- Asset Diversity: The study is limited to WETH-USDC. Future work is needed to see if these results hold for "long-tail" (less liquid) tokens.
- Adversarial Ordering: The framework doesn't fully account for how block builders might maliciously reorder transactions to the user's detriment.
- Expansion to Other Models: The authors plan to apply this methodology to other types of auctions, such as CoW Swap (batch auctions) and MEV Share (rebate systems).
Summary updated