Dataset and scripts

Data and analysis scripts for my master thesis at TU Wien on Ethereum's account abstraction.

Ethereum mainnet data on EIP-7702 and ERC-4337 accounts. First and last block in the data:

• EIP-7702: blocks 22428093 to 24358289 (2025-05-07 to 2026-01-31)
• ERC-4337: blocks 16494438 to 24358289 (2023-01-27 to 2026-01-31)

Sources: Google BigQuery, Alchemy

Files

• parquet_eip7702.zip – EIP-7702 tables
• parquet_erc4337.zip – ERC-4337 tables
• schema_and_load_sql.zip – schema.sql and load.sql for loading everything into DuckDB
• masterthesis_scripts.zip – Scripts for data collection and analysis: BigQuery SQL queries, authorization list fetcher, EIP-7702 authority recovery, and the adoption analysis scripts for Chapter 6 of the thesis

DuckDB replaced the digits in the file names with underscores, so traces_eip____.parquet is the table traces_eip7702 and traces_erc____.parquet is traces_erc4337.

smart_accounts_eip7702 has one row per authorization (1,643,240 distinct accounts).

How to use it

Unzip parquet_eip7702.zip, parquet_erc4337.zip, and schema_and_load_sql.zip into the same folder. It needs about 38 GB.

The easiest way is to query the files directly, without importing:

SELECT * FROM 'path/to/folder/transactions_eip____.parquet' LIMIT 10;

To load everything into a DuckDB database, open a database file first and then import. This needs several hundred GB of free disk space (the imported database is at least 370 GB).

duckdb ~/masterthesis.duckdb

IMPORT DATABASE 'path/to/folder';

Reproducing the analysis

Unzip masterthesis_scripts.zip and follow its README.md, which lists the requirements and the order in which to run the scripts. The scripts read the database from ~/masterthesis.duckdb . Set the environment variable MASTERTHESIS_DB to use a different path.