Simulating Intents

The datasets outlined below are used to analyze the coincidence of wants peer-to-peer by converting on-chain trades to intents, and looking at the price improvement in execution. For each of the multiple datasets utilized in the empirical analysis, the metadata and instructions on how to pull the datasets are outlined below.

Trades

Description

Blockchain trades fetched from Dune Analytics with query trades.dune.sql.

Query sources trades from dex_aggregator.trades, dex.trades, oneinch.swaps and custom query for UniswapX fills. Trades from dex_aggregator.trades and dex.trades are filtered to only ones where sender of the transaction equals protocol's address; with exception of CowSwap, Bebob and 0x Settler.

Each directory contains metadata file describing parameters used when fetching the data: chain, date_from, date_to and tokens.

The dataset is curated by Miha Lotrič, and is part of the TLDR 2025 fellowship program.

Data Format

.parquet

Schema

• id [String]: Trade identifier
• project [String]: Protocol the trade was facilitated through
• block_time [Int64]: Unix timestamp of the block the trade was included in
• amount_usd [Float64]: Trading amount in USD
• token_bought_address [String]: Address of the asset trader received
• token_sold_address [String]: Address of the asset trader gave
• token_bought_amount [Float64]: Fixed-point amount trader received
• token_sold_amount [Float64]: Fixed-point amount trader gave

Collection

Data can be programmatically collected by running the script as shown below:

python -m utils.dune.trades --tokens weth,usdc,wbtc --date-from "2025-03-01 00:00:00" --date-to "2025-03-23 00:00:00" --chain ethereum --performance medium --out-dir data/hist-trades --label ethereum-mar25-1

Parameters:

• --tokens: List of supported token symbols or addresses
• --date-from: Start date in format "YYYY-MM-DD HH:MM:SS"
• --date-to: End date in format "YYYY-MM-DD HH:MM:SS"
• --result-limit: Maximum number of rows to fetch (default: 600,000)
• --chain: Blockchain network (default: ethereum)
• --performance: Query performance (choices: low, medium, high)
• --label: Label for the output directory (default: random UUID)
• --out-dir: Base output directory (default: data)

CowSwap Intents

Description

CowSwap protocol orders fetched from CowSwap's batch API. Only a subset of original parameters is recorded.

File name indicates the chain orders relate to.

Data Format

parquet

Schema

• batch_id [Int64]: Unique CowSwap batch identifier
• uid [String]: Unique CowSwap order identifer
• sell_token [String]: Address of token trader is selling
• buy_token [String]: Address of token trader is buying
• sell_amount [String]: Amount trader is selling (in wei)
• buy_amount [String]: Amount trader is buying (in wei)
• created_at [Int64]: Unix timestamp of order creation
• valid_to [Int64]: Unix timestamp of order expiry
• kind [String]: Whether order is sell or buy
• partially_fillable [Bool]: Whether order is partially fillable
• class [String]: Order class (eg. limit)

Collection

Data can be collected by running utils.dune.cowswap_intent_loader.py. Note that, as of writing, only batches from the last 28 days can be queried.

CowSwap Fills

Description

CowSwap protocol order fills fetched from Dune Analytics with query cowswap_fills.dune.sql. The query sources fills from Dune's dataset gnosis_protocol_v2_multichain.GPv2Settlement_evt_Trade.

Each directory contains metadata file describing parameters used when fetching the data: chain, date_from, date_to and tokens.

Data Format

parquet

Schema

• tx_hash [String]: Transaction hash in which order was filled
• block_time [Int64]: Unix timestamp of the block fill was included in
• orderUid [String]: Unique CowSwap order identifer for filled order
• token_sold [String]: Address of token trader sold
• token_bought [String]: Address of token trader bought
• amount_sold [String]: Amount trader bought (in wei)
• amount_bought [String]: Amount trader sold (in wei)

Collection

Data can be programmatically collected by running the script as shown below:

python -m utils.dune.cowswap_fills --tokens weth,usdc,wbtc --date-from "2024-12-18" --date-to "2025-04-27" --chain ethereum --out-dir data/intents/cowswap/fills --label ethereum_20241218_20250427

Fusion Intents

Description

OneInch Fusion order data streamed from OneInch Fusion websocket feed.

File name indicates the chain order data relates to.

Data Format

jsonl

Schema

See OneInch Fusion docs.

Collection

TBD

Fusion Fills

Description

OneInch Fusion order fills fetched from Dune Analytics with query fusion_fills.dune.sql. The query sources fills from Dune's dataset oneinch.swaps.

Each directory contains metadata file describing parameters used when fetching the data: chain, date_from, date_to and tokens.

Data Format

parquet

Schema

• tx_hash [String]: Transaction hash in which order was filled
• block_time [Int64]: Unix timestamp of the block fill was included in
• order_hash [String]: Unique Fusion order identifer for filled order
• amount_usd [Float64]: Trading amount in USD
• token_bought_address [String]: Address of the asset trader received
• token_sold_address [String]: Address of the asset trader gave
• token_bought_amount [Float64]: Amount trader received (in wei)
• token_sold_amount [Float64]: Amount trader gave (in wei)

Collection

Data can be programmatically collected by running the script as shown below:

python -m utils.dune.fusion_fills --tokens weth,usdc,wbtc --date-from "2024-12-18" --date-to "2025-04-27" --chain ethereum --out-dir data/intents/fusion/fills --label ethereum_20241218_20250427

Prices

Description

Token price candles fetched from Dune Analytics using query prices.dune.sql. Query sources prices from Dune's table prices.usd_daily.

Each directory contains metadata file describing parameters used when fetching the data: chain, date_from, date_to and tokens.

Data Format

parquet

Schema

• day [String]: Date of the candle
• token [String]: Token address
• decimals [Int64]: Token decimals
• price [Float64]: Average price
• price_high [Float64]: Highest daily price
• price_low [Float64]: Lowest daily price

Collection

Data can be programmatically collected by running the script as shown below:

python -m utils.dune.prices --tokens weth,usdc,wbtc --date-from "2024-12-18" --date-to "2025-04-27" --chain ethereum --out-dir data/prices --label ethereum_20241218_20250427

Normalized Intents

TBD