README.md

---
license: odc-by
pretty_name: "Ethereum DEX swap quotes, routing, and perpetual order-book depth"
tags:
  - dex
  - ethereum
  - defi
  - execution-cost
  - slippage
  - routing
  - perpetual-futures
  - order-book
  - trading
  - time-series
task_categories:
  - time-series-forecasting
size_categories:
  - 10K<n<100K
---

# Crypto execution costs

What it costs to trade, quoted at the moment someone would have traded.

Executed trades are on chain forever. Quotes for trades nobody placed are computed on demand and
kept by nobody, so they only exist if they were recorded when they were served.

## Contents

| name | one row is |
|---|---|
| `e10_quote_benchmark` | one swap quote at a fixed size, with the provider fee separated out |
| `e16_dex_routes` | one leg of a route a router chose: venue, pool, amount, and how many venues the trade was split across |
| `e17_perp_depth` | one order book snapshot: spread, level counts, resting notional within fixed distances of mid |

## Reading it

Quotes are what a provider served, which is not what you would necessarily have filled. Provider
fees are separated from price so that a fee policy is not mistaken for routing quality.

`e16_dex_routes` maps where routable liquidity sits rather than where volume went, and it
surfaces venues too small to appear in volume rankings. The split widens sharply with size: one
observation put the same pair through a single venue at small size and ten venues at large size.

`e17_perp_depth` stores book shape rather than raw ladders, as cumulative resting notional within
a band of mid. Where a spread is wider than a band, that band is correctly zero, which happens
often on thinner markets.

## Before you build on this

- Not every provider answers every request. Refusals and rate-limit skips are written as explicit
  error rows rather than omitted, so a provider that was unavailable is distinguishable from one
  that had nothing to offer. Check the `error` column before treating absence as meaning.
- One provider is queried on a subset of pairs. Its coverage is deliberately narrower than the
  others and should not be read as a market-wide view.
- Only some routers expose a leg breakdown. Rows from the others carry the venue name alone,
  which is why `pool` and `swap_amount_raw` are frequently null.
- Depth rows are snapshots at the observation interval, not a tick-level book. A move that
  reverses between observations is invisible.
- Depth covers a single venue, so it sizes that one book rather than the market. Treat it as
  one participant's view; a second venue would be a control and there is not one here.

Partitions are parquet, one file per collection window, under `dataset/YYYY/MM/`. Every dataset here carries a FIXED 7-day sample WINDOW starting at its own first day of collection, together spanning 2026-08-25 to 2026-09-03, so you can check schema, coverage and quality before asking for more. It does not advance, so there is nothing to gain by re-downloading it. The full history is held privately, available on request.

```python
from huggingface_hub import snapshot_download
import pandas as pd, glob

path = snapshot_download("dataforge-labs/crypto-execution-costs", repo_type="dataset",
                         allow_patterns="e10_quote_benchmark/**")
df = pd.concat(map(pd.read_parquet,
                   glob.glob(f"{path}/e10_quote_benchmark/**/*.parquet", recursive=True)))
```

## Coverage

`e0_run_manifest` lists every collection window with its poll counts and failure counts, and is
published in full rather than windowed. Gaps between windows are real, cannot be filled in
afterwards, and nothing here is interpolated.

## License and contact

ODC-BY: use it freely, credit "DataForge (dataforge-labs)". Questions and requests for the full
history via the discussions tab.