Becoming Immutable: How Ethereum is Made
Andrea Canidio, Vabuk Pahari · 2025
Read Becoming Immutable: How Ethereum is Made by Andrea Canidio, Vabuk Pahari (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-04940.txt.
Introduction
Andrea Canidio and Vabuk Pahari peer into the hidden auction that decides which Ethereum transactions land in a block. Standard chain data show only winning blocks, so the routing scramble stays invisible. Their dataset holds 15,097 proposed blocks that lost, a rare look at the cutting room floor. In those blocks, 21% of user trades appear yet never made the chain, which proves routing delays real users.
Execution quality swung widely between rival blocks for the very same swap. Two arbitrage bots anchor a closer test: user swaps moving with the bots executed less often at worse prices. Block makeup, not just fees, steers outcomes.
Catalog record
| CREATOR | Andrea Canidio, Vabuk Pahari |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 81 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-04940
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Becoming Immutable: How Ethereum is Made Thanks: We are grateful to Agnostic Relay for providing the data, especially to Ombre for patiently explaining how relays work. We are also grateful to Agostino Capponi, Hanna Halaburda, Fahad Saleh, and the participants in the workshop “(Back to) The Future(s) of Money II” for their comments and suggestions. Andrea Canidio, Vabuk Pahari Abstract We analyze blocks proposed for inclusion in the Ethereum blockchain during 8 minutes on December 3rd, 2024. Our dataset comprises 38 winning blocks, 15,097 proposed blocks, 10,793 unique transactions, and 2,380,014 transaction-block pairings. We find that exclusive transactions —transactions present only in blocks proposed by a single builder — account for 85% of the fees paid by all transactions included in winning blocks. We also find that a surprisingly large number of user transactions are delayed : although proposed during a bidding cycle, they are not included in the corresponding winning block. Many such delayed transactions are exclusive to a losing builder. We also identify two arbitrage bots trading between decentralized (DEX) and centralized exchanges (CEX). By examining their bidding dynamics, we estimate that the implied price at which these bots trade USDC/WETH and USDT/WETH on CEXes is between 3.4 and 4.2 basis points better than the contemporaneous price reported on Binance. 1 Introduction inline inline todo: inline before submitting, we should make sure that we are consistent with the naming of builders and relays inline inline todo: inline change ”block value” to ”block’s potential revenues” (or somethign else) Blockchains are often described as immutable public ledgers because transactions included in the canonical chain cannot be altered. This static characterization, however, conceals the probabilistic and competitive process that determines which transactions become immutable. In Ethereum, for each block added to the canonical chain, up to 5,000 candidate blocks are constructed, propagated, but ultimately discarded along with their transactions. Each of these discarded blocks represents a counterfactual, unrealized history. In one candidate block, a user might secure a highly profitable trade; in another, the same user could suffer a loss due to an attack. When such differences arise, the choice of winning block has significant consequences for users. Conversely, if blocks differ only in a handful of arbitrage transactions, the outcome may matter little to regular users. Understanding how proposed blocks differ and the resulting implications for users is essential to understanding a system that facilitates roughly USD 5 billion in daily financial exchanges. In this paper, we perform the first analysis of transactions included in non-winning Ethereum blocks. We collect 15,097 non-winning blocks submitted between Ethereum blocks 21,322,622 to 21,322,660 (38 winning blocks), approximately from 2:37:35 PM to 2:45:25 PM UTC on December 3, 2024. Our dataset contains 10,793 individual transactions (identified by their hash) and 2.380.014 individual transactions/blocks combinations, which will be our main unit of analysis. During the approximately eight-minute we consider, the price of ETH against USDC on Binance showed notable volatility. It began at 3,517.45 (2:37:35 PM UTC), rose by 1.06% to 3,554.92 (2:42:12 PM), dipped by 0.37% to 3,541.68 (2:43:13 PM), and then increased again by 0.60% to close at 3,562.79 (2:45:25 PM) (see Figure 1 ). Before presenting our main results, we review the process by which blocks are submitted and selected for inclusion (readers familiar with it may skip this section) and describe our dataset in greater detail, including its limitations. After presenting our findings, we discuss their relevance to the broader debate about Ethereum’s current design and their relation to the existing academic literature. Figure 1 : ETH/USDC price on Binance during our study period 1.1 Background: the journey of an Ethereum transaction On Ethereum, the right to add a new block to the canonical chain is assigned to a proposer (or validator ), randomly selected from addresses that have locked (or staked) a sufficient amount of ETH, Ethereum’s native token. The proposer can assemble the block using transactions from the public mempool, a repository of pending transactions submitted by Ethereum users. However, in over 90% of cases, the proposer delegates block creation to specialized entities called builders , who gather both user transactions and those submitted by searchers . Searchers are automated bots that generate transactions in response to market conditions (for example, price fluctuations) or in response to other pending transactions in the mempool. For builders, selecting which transactions to include and their order is a complex optimization problem: block space is limited, not all transactions can be included, and some may conflict with others. Users and searchers interact with builders in different ways. Some users submit their transactions to the Ethereum public mempool, where builders observe them. However, this exposes them to potential attacks. 1 1 1 For example, if a user swaps assets on a blockchain-based financial market (an Automated Market Maker, or AMM), a malicious searcher can execute a “sandwich” attack: it front-runs the user’s swap by making the same trade first, then back-runs it with the opposite swap. This allows the attacker to buy low and sell high, forcing the victim to trade at worse prices. To mitigate this risk, users may instead use private mempools, services that connect them directly with multiple builders. Searchers always submit their transactions privately, either through known private mempools or via direct communication with one or more builders. Every transaction includes either a fee or a payment to the “Fee Recipient” address, a block-level variable set by the builder (typically an Ethereum address they control). As a result, builders earn fees from users and searchers whenever they include their transactions in a block that is added to the blockchain. The relation between builders and proposers is facilitated by relays . Their primary role is to prevent proposers from altering the submitted blocks, for example, by changing the ‘‘Fee Recipient’’ variable. 2 2 2 Technically, this is achieved by requiring proposers to sign an empty block, which is then completed by the relay. For more details, see https://docs.flashbots.net/flashbots-mev-boost/relay . Once a builder creates a block, it submits it to a relay along with a proposed payment for the proposer. The relay then forwards the highest-paying block and its corresponding payment to the proposer upon request. This process resembles an ascending price auction because the relay continuously broadcasts the value of the highest-paying bid—though this broadcast is observed with a delay due to network latency. An auction cycle lasts approximately 12 seconds, after which the winning block is selected, and a new auction begins. During each auction cycle, builders continuously resubmit blocks, modifying the transactions included and bid amounts. Latency plays a crucial role, as proposers are geographically dispersed, and arbitrage transactions often depend on rapid price movements in traditional “off-chain” financial markets. To account for geographic dispersion and fast-moving arbitrage opportunities, multiple relays operate worldwide, with builders submitting to several at once. As a result, thousands of blocks are created and discarded for every block added to the Ethereum blockchain. 1.2 Our dataset Our primary dataset includes all blocks (winning and non-winning) submitted via a relay called ‘‘Agnostic relay’’ during the study period. 3 3 3 Founded in 2022, Agnostic relay held 20% of the relay market in 2023, declining to around 5% during our study period. We complement our primary