---
language:
- en
license: cc-by-nc-4.0
pretty_name: "Ethereum On-Chain Signals - Free Sample (10K rows)"
size_categories:
- 1K<n<10K
tags:
- ethereum
- blockchain
- defi
- web3
- signals
- on-chain-analytics
- erc20
- erc721
- nft
- dex
- oracle
- mev
- parquet
task_categories:
- tabular-classification
- time-series-forecasting
configs:
- config_name: default
  data_files:
  - split: train
    path: data/sepolia-sample-mixed.parquet
---

# Ethereum On-Chain Signals - Free Sample

**10,000 decoded on-chain events from Ethereum Sepolia testnet**, extracted directly from the P2P network using [Fused Semantic Execution (FSE)](https://deltazerolabs.dev) - a zero-RPC, zero-full-node signal extraction engine.

> **This is a free sample.** The full dataset covers **76M+ signals across 12M+ blocks** on Sepolia alone. Mainnet extraction is underway (genesis-to-tip). Full datasets available at [deltazerolabs.dev](https://deltazerolabs.dev).

---

## What\'s In This Sample

| Signal Type | Count | Description |
|------------|-------|-------------|
| `ERC721_Transfer` | 9,132 | NFT transfers (also includes ERC20 transfers - same topic0) |
| `ERC20_Approval` | 707 | Token spend approvals |
| `WETH_Deposit` | 55 | ETH to WETH wrapping events |
| `Chainlink_AnswerUpdated` | 48 | Oracle price feed updates |
| `Arb_Swap_V2` | 17 | Uniswap V2 swaps |
| `Arb_Sync_V2` | 14 | Uniswap V2 reserve syncs |
| `Arb_Swap_V3` | 10 | Uniswap V3 swaps |
| `Bridge_LayerZero_Packet` | 7 | Cross-chain bridge messages |
| `WETH_Withdrawal` | 7 | WETH to ETH unwrapping |
| `Bridge_Wormhole_LogMessagePublished` | 2 | Wormhole bridge events |
| `DAO_VoteCast` | 1 | Governance vote |

**Block range:** 1,394,316 to 7,700,283 | **25 columns** | **1.5 MB parquet**

---

## Schema

```
block_number        int64       Block height
block_hash          string      Block hash
timestamp           int64       Unix timestamp
tx_hash             string      Transaction hash
tx_index            int64       Position in block
from_address        string      Transaction sender
to_address          string      Transaction recipient
gas_price           int64       Gas price (wei)
gas_used            int64       Gas consumed
value               int64       ETH value transferred (wei)
log_index           int64       Log position in receipt
contract_address    string      Emitting contract
signal_type         string      Classified event type
topic0              string      Event signature hash
topic1-topic3       string      Indexed event parameters
data                string      Non-indexed event data (hex)
decoded_data        string      Human-readable decoded parameters
chain_id            int64       Chain identifier
created_at          string      Extraction timestamp
```

---

## Quick Start

```python
import pandas as pd

# Load directly from Hugging Face
df = pd.read_parquet("hf://datasets/MikeKuykendall/ethereum-signals-sample/data/sepolia-sample-mixed.parquet")

# All Uniswap V2 swaps
swaps = df[df["signal_type"] == "Arb_Swap_V2"]

# Oracle price updates
oracles = df[df["signal_type"] == "Chainlink_AnswerUpdated"]

# NFT transfers for a specific contract
nfts = df[(df["signal_type"] == "ERC721_Transfer") & (df["contract_address"] == "0x...")]
```

---

## How This Data Was Extracted

Unlike RPC-based indexers (Dune, The Graph, Alchemy), this data was extracted **directly from Ethereum\'s P2P network** using a custom devp2p client. No RPC rate limits. No API keys. No full node required.

The extraction engine uses **Fused Semantic Execution (FSE)** - a technique that scans raw receipt logs against multiple signal patterns in a single O(1) pass, classifying events as they stream off the wire.

**What makes this different:**
- **Zero RPC dependency** - connects directly to Ethereum peers via devp2p/ETH68
- **Pre-decoded signals** - events are classified and decoded at extraction time
- **Multi-signal in one pass** - DEX swaps, oracle updates, bridge events, governance - all extracted simultaneously
- **Minimal storage** - only signals are kept, raw blocks are discarded

---

## Full Dataset Available

This sample is **0.013%** of the Sepolia dataset. The full extraction includes:

| Product | Signals | Size |
|---------|---------|------|
| NFT Transfers (full) | ~45M events | 10.9 GB |
| ERC20 Approvals (full) | ~8M events | 1.8 GB |
| DEX Activity (full) | ~1.2M events | 172 MB |
| WETH Wrapping (full) | ~900K events | 177 MB |
| Oracle Price Feeds (full) | ~500K events | 95 MB |
| Bridge Events (full) | ~150K events | 24 MB |
| Governance (full) | ~5K events | 935 KB |
| **Total (Sepolia)** | **76M+ signals** | **13.2 GB** |

**Mainnet extraction (genesis to tip) is currently running.**

### [Get the full dataset at deltazerolabs.dev](https://deltazerolabs.dev)

---

## Use Cases

- **MEV Research** - Cross-reference swap events with mempool data to study arbitrage patterns
- **DeFi Analytics** - Track oracle updates, liquidity events, and bridge flows across time
- **ML Training** - Labeled on-chain events ready for time-series models and anomaly detection
- **Academic Research** - CC-BY-NC licensed, citation-friendly, parquet format

---

## License

This sample is released under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/). Free for research, education, and non-commercial use. For commercial licensing of the full dataset, visit [deltazerolabs.dev](https://deltazerolabs.dev).

---

## Citation

```bibtex
@dataset{kuykendall2026ethereum,
  title={Ethereum On-Chain Signals Dataset},
  author={Kuykendall, Michael Allen},
  year={2026},
  publisher={Delta Zero Labs},
  url={https://deltazerolabs.dev}
}
```

---

**Built by [Delta Zero Labs](https://deltazerolabs.dev)** | *Extracting alpha from the wire.*

[Website](https://deltazerolabs.dev) | [Full Dataset Pricing](https://deltazerolabs.dev/#pricing) | [Contact](mailto:mike@deltazerolabs.dev)