bitcoin-data

Collection of public Bitcoin related datasets.

Blocks #

Datasets about Bitcoin blocks and block headers.

Bitcoin blockchain fork measurements (KIT DSN)

Observed blockchain forks as plain-text tables, one row per fork with the height, the reception times of the main-chain and orphaned block, their delay, how many peers announced the earlier one, and each block's inferred miner. Recorded by KIT's Bitcoin monitoring nodes, with a per-block file listing each peer announcement of the competing blocks. Useful for studying fork resolution and miner advantage; miner labels come from a block explorer.

Bitcoin mining evidence from merge-mined chains

Merge Mining Research publishes Bitcoin parent headers and, where available, coinbase data recovered from merge-mined chains such as Namecoin. The datasets show which child chains shared a Bitcoin block's proof of work and provide evidence for pool attribution. They cover blocks on Bitcoin's active chain, stale header candidates and documented consensus-invalid blocks. The checks on stale candidates do not establish full-block validity. Merge Mining Monitor combines these historical datasets with live observations from Bitcoin Core and active child chains.

Invalid blocks

Bitcoin blocks with valid proof of work that fail a named consensus rule, one JSON record per header with the failure reason, evidence and observations. Records come from the P2P network, merge-mined chains and archived block explorers, and CI verifies each record's proof of work and evidence. Useful for studying consensus failures and historical incidents; blocks only reported as invalid are kept in a separate file.

  • Category: Blocks
  • Format: JSON Lines, JSON, Binary
  • Period: 2010-08-15 – present
  • License: CC0 1.0 (data), MIT (code)
  • Author: 0xB10C and contributors
  • Tags:

Merged mining fork analysis

JSON lists of Bitcoin blocks tagged with merge-mined child chains, difficulty epochs, stale blocks recovered from Namecoin merged-mining data, and stale block snapshots from fork monitors and explorers. Compiled by Stifter et al. for the paper Echoes of the Past from coin clients, the KIT and Decker-Wattenhofer monitors and explorer APIs, with a Litecoin counterpart. Useful for comparing live fork monitoring with merged-mining stale block recovery; the release is partial and large files are tarballs.

  • Category: Blocks
  • Format: JSON, Binary
  • License: Not specified
  • Author: Nicholas Stifter, Philipp Schindler, Aljosha Judmayer, Alexei Zamyatin, Andreas Kern and Edgar Weippl
  • Tags:

Open Bitcoin Metrics

Daily Bitcoin on-chain series covering block production, blockspace, fees, miner revenue, difficulty, estimated hashrate, supply, UTXO counts and coin-age measures like Bitcoin days destroyed and liveliness. Each series is a CSV with date, identifier, value and unit, computed from a Bitcoin Core full node and a spent-output indexer. Useful for reproducible econometric work, since every series ships its definition and the script computing it; each archived release is a snapshot.

Stale blocks

Stale Bitcoin blocks and headers observed on the network, one CSV row per block with height, hash and raw header, plus full blocks where available. Node operators contribute their chain tips from Bitcoin Core's getchaintips RPC, supplemented by fork monitors and block explorer archives. Useful for stale-block rates and propagation studies; listed blocks are not verified to be fully consensus-valid.

Transactions #

Datasets about Bitcoin transactions and the mempool.

Bitcoin address-transaction bipartite graph

Bipartite graph of Bitcoin addresses and transactions as a text edge list (address, transaction, input/output flag, satoshi value) and as binary CSR arrays with a small C++ reader. Built at UMass Lowell by parsing Bitcoin Core block files with rusty-blockparser and numbering every address and transaction as a vertex. Useful for large-scale analysis of address reuse and anonymity; it is a fixed snapshot and vertex numbers are not mapped back to addresses or transaction ids.

  • Category: Transactions
  • Format: Plain text, Binary
  • Period: 2009-01-03 – 2018-06-07
  • License: CC BY 4.0
  • Author: Anil Gaihre, Yan Luo, Hang Liu (UMass Lowell)
  • Tags:

Bitcoin addresses and their categories

One CSV row per Bitcoin address with a category label: mining pool, miner, CoinJoin-like, gambling, exchange or services. Compiled at Wrocław University of Science and Technology from plausible assumptions and third-party labelling services, to train and validate supervised classifiers for address types. Useful as a label set for address-classification experiments; the labels are heuristic ground truth rather than verified attribution, so some are wrong.

Bitcoin blockchain data by cohort analysis

CSV tables of daily Bitcoin output cohorts, one row per date with counts of unspent outputs bucketed by age and of spent outputs bucketed by lifespan, plus metadata tables describing the columns. Computed at Duke Kunshan University by querying the transaction input and output history and aggregating it per day. Useful for work on holding behaviour without reprocessing the chain; only some files cover Bitcoin, the rest cover other coins.

  • Category: Transactions
  • Format: CSV
  • Period: 2009-01-03 – 2021-02-10
  • License: CC BY 4.0, per the Dataverse terms of use; no license block set
  • Author: Yulin Liu, Luyao Zhang and Yinhong Zhao (Duke Kunshan University)
  • DOI: 10.7910/DVN/XSZQWP
  • Tags:

Bitcoin blockchain in BigQuery

Two BigQuery tables covering the chain, one row per block and one per transaction, with each transaction's inputs and outputs nested alongside their addresses, script, type and value, plus views that flatten them. Google's public dataset program hosts it, loaded by the open source Blockchain ETL pipeline from a node. Useful for SQL analysis of the whole chain without running an indexer; querying needs a Google Cloud project.

  • Category: Transactions
  • Format: CSV, JSON Lines, Parquet
  • Period: 2009-01-03 – present
  • License: Not specified (bitcoin-etl extraction code is MIT)
  • Author: Google Cloud Public Dataset Program
  • Tags:

Bitcoin blockchain optimized for machine learning prediction model

Separate block and transaction CSV tables, joined on block hash: blocks carry size, height and creation time, transactions carry fee, size, broadcast time and waiting time until inclusion. Sampled from blockchain.com APIs and a Bitcoin Core node by Enrico Tedeschi at UiT for work on predicting which transactions miners include. A README documents every column; coverage is limited to the sampled monthly windows, and transaction latency can be negative.

Bitcoin dust transactions

Bitcoin transactions creating or spending dust outputs, one text line per transaction with block time, height, fee, size and each input and output as address identifier, amount and script type, plus CSV address mappings and entity labels. Extracted from the blockchain at the University of Pisa; a companion record holds all transactions. Useful for studying dust creation, consumption and dust attacks; dust is a fixed satoshi threshold.

Bitcoin memory pool transaction capture

One row per transaction in a node's mempool: txid and wtxid, size, weight, ancestor and descendant fees, entry time, and the decoded inputs and outputs. Captured by the authors at Tianjin University of Technology from their own full node and published as a SQL Server dump. Useful for mempool and fee modelling; it is one node's view, and height is the chain tip at entry, not a confirmation height.

Bitcoin mempool transaction lifecycle

One Parquet row per observed transaction with first-seen time, time pending, fee rate, ancestor package fields and an outcome of confirmed, dropped or unresolved, plus tables comparing pending sets across providers and summarising block composition. A collector polls several mempool providers and records each transaction until it confirms or disappears. Useful for fee and confirmation-time modelling; the public release is a sample window and fee fields are often missing.

Bitcoin OP_RETURN metadata

One CSV row per transaction carrying an OP_RETURN output, giving the transaction hash, its time, the payload and the protocol it is attributed to. Built at the University of Cagliari by scanning the blockchain for OP_RETURN scripts and matching them against known data-embedding protocols. Useful for studying what is stored on chain and by which application; only protocols the authors identified are labelled, and the scan covers a fixed prefix of the chain.

Bitcoin transaction graphs: training, labelled entities, and CoinJoins

Address-transaction graphs for machine learning on Bitcoin: training graphs with CSV node features, edge features and heuristic clustering labels, plus per-transaction subgraphs with known entity labels and CoinJoin subgraphs as Parquet tables with JSON metadata. Built by the authors from blockchain data and address-clustering heuristics as companion data for a graph neural network clustering method. Useful for training and benchmarking address clustering; heuristic labels may wrongly merge distinct users.

Bitcoin transaction network (Antulov-Fantulin et al.)

A Bitcoin transaction network edge list, one comma-separated line per transfer with integer source and target node ids, a Unix timestamp and an amount in bitcoin. Extracted from the blockchain by the authors of a complex-networks paper on inferring short-term volatility, archived on Zenodo as one tar archive. Useful for temporal network analysis; node ids are opaque, and the volatility file named in the record description is missing.

Bitcoin transaction network (Kondor et al.)

Tables of the Bitcoin transaction graph with one row per transaction, input, output and block, where addresses and transaction hashes are replaced by numeric identifiers mapped in separate files, as compressed CSV and plain-text tables. Extracted from the blockchain with a modified bitcoind client by researchers at Eötvös Loránd University. Useful for network-science studies of payment flows and wealth; identifiers differ between releases, so their files cannot be mixed.

Bitcoin transaction OP_RETURN messages

One row per OP_RETURN output: its script in assembly, hex and ASCII, its index and value, and the transaction and block it sits in. Extracted from the blockchain by the authors at Tianjin University of Technology and published as a SQL Server dump. Useful for studying data embedded in Bitcoin; the input value column is empty, leaving fees negative, and the dump covers fewer blocks than the record claims.

Bitcoin transactions (full transaction list)

All Bitcoin transactions up to a given block height, one compressed text line per transaction with block time, height, fee, size and each input and output as a numeric identifier, amount and script type. Extracted at the University of Pisa as the full-chain companion of the authors' dust record. Useful for whole-chain graph and clustering analysis; compressed CSV maps are needed to resolve identifiers to addresses and transaction hashes.

Bitcoin transactions from 2009 to 2020

Bitcoin value transfers, one row per sender and recipient pair with a Unix timestamp, the sending address, the receiving address and the amount in satoshi. Deposited on Harvard Dataverse as CSV files split by calendar month, some also served as tab-separated tables. A simple starting point for transaction-graph work, but the deposit does not document how the transfers were extracted, so their exact semantics are unclear.

Bitcoin UTXO chainstate and transaction tables

The unspent output set as a SQLite table, one row per output with txid, index, value, script type, height and coinbase flag, plus a CSV table of one row per transaction with sizes, input and output counts and input txids. Extracted at INAOE from Bitcoin Core, the output table read from its LevelDB chainstate. Useful for storage-scalability work; both tables are fixed snapshots split into parts that must be rejoined.

Bitcoin UTXO value vectors

Per-height archives of the unspent output set: the exact vector of positive output values as 64-bit integers, sampled fixed-size output records, and the values created and spent in the window of blocks ending at each height. Reconstructed by replaying the active chain over Bitcoin Core's RPC for a power-law study of UTXO values. Useful for value-distribution work; each archive is a snapshot at one height, not a continuous history.

BitcoinTemporalGraph

A temporal directed graph of value transfers between clustered Bitcoin entities as a PostgreSQL dump, with per-entity degree, amount and cluster statistics and per-pair transfer counts and block heights. Built by the authors from the blockchain with address clustering and published alongside graph-learning baselines. A companion CSV labels addresses with an entity name and type, but those labels are scraped from forum posts and are weaker than the graph.

Change output ground truth

One CSV row per Bitcoin transaction with a known change output, giving the transaction hash and that output's index. Malte Möser and Arvind Narayanan derived it from the blockchain by extracting change outputs revealed by the multi-input heuristic, as companion data to their address clustering paper. It serves as a labelled set for evaluating change detection heuristics, but the labels come from a heuristic and inherit its mistakes.

  • Category: Transactions
  • Format: CSV
  • Period: until 2021-06
  • License: Not specified
  • Author: Malte Möser and Arvind Narayanan (Princeton University)
  • Tags:

CoinJoin input-output mappings

CoinJoin transactions as JSON, each input and output carrying address, value and owning wallet, plus a CSV counting the mappings enumerated per transaction. Produced by the CRoCS group at Masaryk University from simulated Wasabi 2 rounds with known ownership, plus a few real mainnet transactions. Useful for measuring how much ambiguity arbitrary-value CoinJoins create; the repository documentation is minimal, so the file layout must be read from the code.

  • Category: Transactions
  • Format: JSON, CSV
  • Period: 2025-04-10 – 2025-04-17
  • License: MIT
  • Author: Jiri Gavenda, Petr Svenda, Stanislav Bobon and Vladimir Sedlacek (CRoCS, Masaryk University)
  • Tags:

Dumplings CoinJoin extraction scan

CoinJoin transactions of several wallet designs found by the Dumplings scanner, one text line per transaction with txid, block hash, block time, inputs and outputs, plus curated false positives, missed transactions and a JSON txid-to-coordinator map. Produced by the CRoCS group at Masaryk University by scanning a Bitcoin Core node with their Dumplings fork. Useful for studying CoinJoin volumes and usage; detection is heuristic, so false positives and misses occur.

Fee estimates analysis

JSON records of fee rate estimates logged every minute by a Bitcoin Core node, pairing a mempool-based next-block estimate and estimatesmartfee with feerate percentiles of the block mined next. Collected by Abubakar Sadiq Ismail from a patched node's trace and debug logs; a separate branch covers several block-template forecasters. Useful for scoring estimators against confirmed blocks; fee rates are strings in sat/kvB and each branch is a short sample from one node's view.

GraphSense public TagPacks

Attribution tags for cryptoasset addresses as YAML TagPacks: a record gives an address a label, category, actor and link to its public source, in a pack with creator, confidence and last-modified date. Curated by the GraphSense team and contributors, who may submit only tags from public, dereferenceable sources, with a separate ActorPack listing real-world entities. Useful as ground truth for address clustering; some packs are derived snapshots of third-party services.

  • Category: Transactions
  • Format: Plain text
  • Period: ongoing
  • License: MIT
  • Author: Bernhard Haslhofer, Melitta Dragaschnig, Rainer Stütz, Matteo Romiti and Gibran Gomez (GraphSense)
  • Tags:

Mempool cluster linearization artefacts

Bitcoin transaction clusters, one hex-serialised cluster per line of a compressed text file, with each transaction's fee, size and dependencies. One set comes from replaying recorded Bitcoin network traffic through a cluster-mempool build of Bitcoin Core; the rest are randomly generated tree, medium-density and bipartite clusters. Benchmark code and result logs are included; the replayed clusters follow the build's cluster limits, not those enforced when the traffic was recorded.

  • Category: Transactions
  • Format: Plain text
  • License: Not specified (data), MIT per source file headers (code)
  • Author: Arman Mollakhani, Pieter Wuille and Dongning Guo
  • Tags:

Mempool snapshots and accelerator prices

One row per unconfirmed transaction in a full mempool dump, with fee, size, ancestor and descendant fields and dependencies, plus accelerator price quotes for those ids. Researchers at MPI-SWS dumped getrawmempool from their node and queried a mining pool's acceleration service. Accelerator prices cannot be reconstructed from chain data, making this a rare record of what such services charge; coverage is limited to a single provider.

  • Category: Transactions
  • Format: CSV, JSON
  • Period: 2020-11-24 – 2020-12-01
  • License: Not specified
  • Author: Johnnatan Messias, Vabuk Pahari, Balakrishnan Chandrasekaran, Krishna P. Gummadi and Patrick Loiseau (MPI-SWS)
  • Tags:

Mempool transaction inter-arrival times

One tab-separated row per transaction entering a Bitcoin mempool: txid, serialized size, input and output counts, their ratio, millisecond arrival timestamp and the interval since the previous transaction. Collected by the author at the University of Oslo for work on transaction features and synthetic workload generation for UTXO-based ledgers. Useful for modelling arrival processes and size distributions; the record has no README, so columns are interpreted from the header alone, and it reflects a single node's view.

Milk Sad weak key addresses

Bitcoin addresses of wallets whose keys came from a broken random number generator or a guessable passphrase, sorted text files, one address per line, split by generator and key derivation. The Milk Sad researchers brute-force the seed space of each flaw and keep addresses that a bloom filter shows were used on-chain. Useful for flagging affected wallets and thefts from them; publication is selective, so a missing address proves nothing.

  • Category: Transactions
  • Format: Plain text
  • License: Not specified (the repository states that a license, possibly CC BY 4.0, is still under consideration)
  • Author: Milk Sad research team
  • Tags:

Non-standard transactions

Mined Bitcoin transactions that violate Bitcoin Core's standardness policy, one CSV row per transaction with height, mining pool, reason, txid, size and fee details. Blocks are replayed through a Bitcoin Core test node to find transactions its policy would reject. Useful for studying which pools mine non-standard transactions and why; results depend on the policy of the Bitcoin Core version used.

OFAC sanctioned digital currency addresses

Every digital currency address designated by the US Office of Foreign Asset Control, as one address per line in a text file and as a JSON list, with one pair of files per asset. A script extracts them from the Treasury's SDN XML, and a workflow regenerates the lists. Useful for checking whether an address is sanctioned; the files hold current designations only, with earlier ones in the git history.

  • Category: Transactions
  • Format: Plain text, JSON
  • Period: ongoing
  • License: MIT (tool); the addresses come from the US Treasury SDN list
  • Author: 0xB10C
  • Tags:

ORBITAAL temporal Bitcoin entity transaction graph

Entity-level Bitcoin transfers as a temporal graph: a stream graph with one Parquet row per entity-to-entity transfer carrying block timestamp and value in satoshi and USD, snapshots at several time resolutions, and node tables with entity balances and addresses. Built from the blockchain with bitcoin-etl, clustering addresses by the common-input heuristic plus WalletExplorer labels. Useful for temporal network analysis; the clustering is heuristic and can merge unrelated users.

Ransomware payments in the Bitcoin ecosystem

Addresses and payments tied to ransomware families, as CSV tables of per-family seed addresses, collector addresses with their multi-input cluster id, and the transactions paying them with hash, timestamp and value. Produced by the authors on the GraphSense platform, which expands the seed addresses into clusters and follows their payments. Useful for estimating ransom revenue per family; the figures are a lower bound and the clustering can merge unrelated users.

Sextortion spam Bitcoin addresses

A plain list of Bitcoin addresses, one per line with no header, each demanded as payment in a sextortion spam email. Extracted from a corpus of spam messages by the authors of the companion paper on sextortion in the Bitcoin ecosystem. Useful as ground truth for clustering and extortion-campaign studies; it carries addresses only, with no amounts, message text or campaign labels.

The Bitcoin brain drain

CSV tables covering Bitcoin brain wallets, whose keys come from a password: transaction inputs and outputs for drained wallets, a brain wallet master table, a block-to-pool mapping and an R analysis script. Built by the authors by guessing candidate passwords from word lists and matching the derived addresses against the chain. Useful for studying brain wallet use and the drainers that empty them; only wallets whose passwords were guessed appear.

Transaction graph dataset for the Bitcoin blockchain

Bitcoin transfers as graph edges, one text line per transfer with a BTC-IN or BTC-OUT marker, block number, position in the block, source, destination and amount, plus a file mapping block numbers to timestamps. Extracted from the mainnet blockchain at Bogazici University for work on fraud detection in transaction graphs. Useful for large-scale graph analysis; the block ranges are split across several Zenodo records of bzip2 archives fetched separately.

Transaction ordering corpus

One semicolon-separated CSV row per mined transaction with its inferred mining pool, position in the block, fee, feerate, child-pays-for-parent flag and pre-joined input values, plus block and miner tables. Researchers at MPI-SWS built it from a full node, resolving spent inputs and inferring each block's miner. It suits studies of transaction ordering and fee behaviour, but miner labels and the CPFP flag are inferred, not chain facts.

  • Category: Transactions
  • Format: CSV
  • Period: 2018-01-01 – 2020-12-31
  • License: Not specified
  • Author: Johnnatan Messias, Mohamed Alzayat, Balakrishnan Chandrasekaran, Krishna P. Gummadi, Patrick Loiseau and Alan Mislove (MPI-SWS)
  • Tags:

Unconfirmed transaction captures

Unconfirmed Bitcoin transactions observed in a mempool, one line per transaction in a daily file with the txid, the decoded transaction and its Bitcoin Core mempool entry, including fee, vsize, dependencies and replaceability. Collected for a paper on address clustering by patched Bitcoin Core nodes in several countries that record transactions as they arrive. Useful for studying replaced and never-confirmed transactions; coverage depends on what those nodes received.

Unindexed Bitcoin transaction chunks for UTXO output classification

One CSV row per transaction output, giving the transaction id, output index, amount, scriptPubKey, block height, transaction size and a storage class label. Extracted at INAOE from Bitcoin Core's chainstate for the Mech-Scal storage scalability work, which sorts outputs into immutable, low-churn and high-churn classes. Useful as input for output classification work; the chunks are a fixed snapshot split across several deposits and hold CSV text despite a .bin extension.

WatchYourBack address tags

One row per Bitcoin address in headerless CSV files, with a currency ticker, a category, an owner tag, an optional subtype and a reference URL. Collected by the IMDEA Software Institute authors of WatchYourBack, largely by aggregating third-party tag sources, explorers and abuse reports. Useful for labelling exchange and malware addresses during graph exploration; one file tags Lightning channel funding addresses, and tag quality depends on the original reporter.

  • Category: Transactions
  • Format: CSV
  • License: GPL-3.0; aggregated third-party tags carry their own terms
  • Author: Gibran Gomez, Pedro Moreno-Sanchez and Juan Caballero (IMDEA Software Institute)
  • Tags:

P2P network #

Datasets about the Bitcoin peer-to-peer network: nodes, addresses, propagation.

achow101 DNS seeder crawler dump

The node database of achow101's Bitcoin DNS seeder as a text dump, one line per known address with reachability over several time windows, last-seen time, height, services, version and user agent. Written continuously by the dnsseedrs crawler behind the seeder, which asks nodes for addresses and keeps retrying each one. Useful for node counts per network and software adoption; a single live snapshot from one vantage point, without history.

ASMap files for Bitcoin Core

Maps from IP address ranges to autonomous system numbers, encoded for Bitcoin Core's asmap option, with one filled and one unfilled binary file per run, organised by year, plus a copy of the latest map at the repository root. Each map is created collaboratively by several participants from RPKI, IRR and routing data using kartograf, and verified and PGP-attested before merging. Useful for diversifying peer selection by network operator rather than by IP prefix and for studying how Bitcoin nodes are distributed across autonomous systems; reproducing a map requires the kartograf version used at the time.

Bitcoin DNS seeder and crawler archive

Answers of Bitcoin's mainnet DNS seeds, one JSON line per query with record type, status, TTL and returned addresses, plus unmodified seeder and crawler dumps archived as gzipped release assets with checksums. GitHub Actions query every seed and service-bit subdomain with dig and download each operator's published dump. Useful for comparing what bootstrapping nodes are told with independent crawler views; answers come from one vantage point, so per-seed counts are lower bounds.

Bitcoin network monitor snapshots (KIT DSN)

Daily snapshots of the peers connected to KIT's Bitcoin monitoring nodes, one JSON file per day with each peer's services, version, user agent, autonomous system, coarse location, relay statistics and latency. The monitor connects to reachable nodes and records what they relay, following the group's published methodology. Useful for studying the node population and propagation over many years; peer IP addresses are anonymised.

Bitcoin node census

One JSON object per crawl, giving reachable node counts with how many support v2 transport and compact block filters, split by connection type, plus the crawl's duration and total nodes contacted. A Rust crawler walks outward from one seed node and appends a line per run. Useful for tracking feature adoption; onion, I2P and CJDNS nodes are not reached, though the schema has buckets for them.

Bitcoin node OS fingerprints

Guessed operating systems of reachable Bitcoin nodes, one CSV per scan with a row per host giving IP address, port, OS class, confidence and host state. The maintainer takes targets from a btcnodes.io snapshot and runs nmap OS detection against each host's Bitcoin port, keeping only coarse OS classes. Useful for the operating-system mix of reachable nodes; per-host guesses are noisy, middleboxes distort fingerprints, and only IPv4 and IPv6 hosts are scanned.

Bitcoin P2P network crawler snapshots (LORIA)

Per-crawl snapshots of the Bitcoin mainnet network: JSON lists of discovered IPv4 nodes (hashed IP, port, services, height), CSV counts of GETADDR responses, and JSON exports of reachable nodes with protocol version, user agent, geolocation and autonomous system. LORIA researchers collected it with a modified bitnodes crawler described in their papers. Useful for studying the reachable node population; IPs are hashed and truncated, and IPv6 and Tor nodes are absent.

  • Category: P2P network
  • Format: JSON, CSV, Binary
  • Period: 2020-02-07 – 2020-07-09
  • License: Not specified (the accompanying papers on Zenodo are CC BY 4.0)
  • Author: Jean-Philippe Eisenbarth, Thibault Cholez, Olivier Perrin (LORIA, Université de Lorraine, Inria)
  • Tags:

bitcoin.fish.foo DNS seeder crawler dump

The node database of Will Clark's Bitcoin DNS seeder seed.bitcoin.fish.foo as a text dump, one line per known address with reachability over several time windows, height, services, version, user agent and retry statistics. Written continuously by a dnsseedrs crawler behind the seeder. Useful as an independent vantage point, since this seeder is not hardcoded in Bitcoin Core; a single live snapshot without history.

Bitnod.es node explorer exports

Exports of reachable Bitcoin nodes from the Bitnod.es node explorer, one CSV file per day with each node's address, port, country, ISP, services, protocol version, user agent and block height, covering IPv4, IPv6, Tor and I2P addresses. The site runs the bitnodes crawler and publishes a cumulative last-seen export per day. Useful for node counts by country, provider and software version; only listening nodes are included.

Block arrival times

Timestamps at which blocks arrived at various Bitcoin nodes, one CSV file per node with height, hash and millisecond arrival time. Contributors extract the timestamps from their Bitcoin Core debug logs with the included script, and CI checks them against the block header time. Useful for measuring block propagation delay; coverage differs by node and period.

Blockchain overlay known-peer crawls

Edge lists of the known-peer graphs of several cryptocurrency networks, including Bitcoin: one file per crawl snapshot, one line per node listing the ids of the peers it reported. Researchers at Cyprus University of Technology and Telefonica built them by asking every reachable node for its addresses. Useful for studying overlay topology and partitioning attacks; the edges are potential connections and IP addresses are replaced by ids.

BTC Nodes (btcnodes.io) reachable node snapshots

Snapshots of reachable Bitcoin nodes from a revival of the bitnodes.io crawler, each a JSON document listing every node with version, user agent, services and height across IPv4, IPv6, Tor, I2P and CJDNS, with monthly archives adding geolocation and ASN. The crawler recursively requests addresses from nodes and publishes snapshots through a rate-limited public API and an archive page. Useful for node counts by network, country and software; only listening nodes are counted.

getrawaddrman exports

Daily exports of the address manager of Bitcoin Core nodes via the getrawaddrman RPC, one JSON file per node with every known address and its metadata, with history in the git log. A scheduled job fetches the exports from the peer.observer nodes and commits them. Useful for studying address propagation and addrman churn; the nodes are a small sample with their own connectivity.

Instrumented Bitcoin Core experiment logs (NISLAB)

Plain-text event logs from Bitcoin Core nodes patched to record peer-to-peer events, each line carrying a human-readable time, a monotonic timestamp and an event tag covering compact block arrival and reconstruction, mempool dumps, orphan transaction handling and scripted node churn. Boston University's NISLAB recorded them for its churn and orphan-transaction papers. Useful for replicating those analyses; each git branch is a separate experiment with its own layout.

  • Category: P2P network
  • Format: Plain text
  • Period: 2020-05-22 – 2020-07-21
  • License: Not specified
  • Author: Muhammad Anas Imtiaz, David Starobinski, Ari Trachtenberg and Nabeel Younis (NISLAB, Boston University)
  • Tags:

Mempool-based transaction auditing recordings

CSV tables of first-seen times for unconfirmed transactions, one row per transaction hash and one column per observing node holding the Unix-second timestamp at which it first received that transaction. The authors recorded them in the mempools of their own nodes in Europe, the United States and Asia for a paper on mempool-based transaction auditing. Useful for studying regional relay delay; unobserved transactions are left empty.

Network propagation timing preview (DataForge)

Parquet tables of block and transaction propagation timing across Bitcoin and several altcoins: one row per block or transaction with a first-seen timestamp, an announcement count and delay percentiles, plus per-window relay fee and client version tables. DataForge records them from its own nodes' peer connections and publishes this repository as a free preview of a paid feed. Useful for comparing propagation across chains; timings come from a single observation point and carry no per-peer detail.

Node count history and user agent breakdown (Luke Dashjr)

The data behind Luke Dashjr's node count charts: a plain-text history with one line per sample giving a Unix timestamp, the listening node count, the estimated unreachable count and counts for selected software and service groups. A companion JSON file breaks the same counts down by user agent and adds per-service-bit shares. Useful for node population trends; how the unreachable estimate is produced is not documented.

P2P network health metrics

Daily time series on the health of Bitcoin's P2P network: reachable nodes by network type, unique and long-term reachable node estimates, and the quality of addresses served by each DNS seed. Produced by virtu's p2p-metrics-exporter, which crawls the network and queries the DNS seeds. Useful for tracking node counts and DNS seed quality over time; only listening nodes are counted.

Transaction inv timing traces with originator ids

Gzip-compressed CSV traces of transaction inv announcements: one row per announcement with a millisecond timestamp, a keyed hash of the announcing peer's address and the txid. A modified bitcoinj crawler recorded them at Ulm University and at cloud hosts in several regions, some runs sharing a key so originator ids match across vantage points. Useful for studying relay timing and origin inference; an announcing peer is not necessarily the originator.

Mining #

Datasets about mining pools, hashrate and coinbase data.

Blocks by mining pool

One CSV row per Bitcoin block with its height, hash and the slug of the pool that mined it, attributing each block individually rather than listing coinbase tags. Labels come from the mempool.space block API, queried by a script in the repository that regenerates the rows. Useful for joining pool identity onto block-level analyses; attribution follows mempool.space's heuristics, and blocks it cannot identify carry the slug unknown.

Cambridge Bitcoin Electricity Consumption Index

One row per day in a CSV, with a Unix timestamp and lower-bound, upper-bound and best-guess estimates of the Bitcoin network's annualised electricity consumption in terawatt-hours. Modelled by the Cambridge Centre for Alternative Finance from hashrate and a profitability-weighted mix of hardware efficiencies. The site plots the same model as power demand in gigawatts; the figures rest on assumptions about hardware mix and data-centre overhead, and are estimates, not measurements.

  • Category: Mining
  • Format: CSV
  • Period: 2010-07-18 – present
  • License: CC BY-NC-SA 4.0
  • Author: Cambridge Centre for Alternative Finance (Cambridge Judge Business School)
  • Tags:

Daily blocks mined per entity

Blocks mined per day by each Bitcoin mining entity, one CSV row per pool or unattributed address with a column per day. Compiled by Edinburgh's Blockchain Technology Lab from BigQuery block data, attributing blocks to pools via coinbase tags and known addresses, for a paper on confidence in mining power estimates. Useful for Nakamoto coefficient and pool-share analyses; attribution is heuristic, and unmatched addresses count as separate entities that may share an operator.

  • Category: Mining
  • Format: CSV
  • Period: 2018-01-01 – 2023-12-31
  • License: MIT
  • Author: Mary Milad, Christina Ovezik, Dimitris Karakostas and Daniel W. Woods (Blockchain Technology Lab, University of Edinburgh)
  • Tags:

Mining pool reward distribution (A Deep Dive into Bitcoin Mining Pools)

Per-block mining pool attributions with coinbase script, payout addresses and attributed pool, address-to-cluster mappings, cluster tags, pool member lists and raw payout flow graphs for AntPool, BTC.com and ViaBTC, as JSON and CSV in a zip archive. The authors derived it from mainnet blockchain data, address clustering and pool tag lists from blockchain.info, BTC.com and Blocktrail. Useful for studying reward distribution within pools; attributions rest on heuristics and third-party tags.

Mining pools

Known Bitcoin mining pools with their coinbase tags and coinbase output addresses, one JSON file per pool, for mainnet and signet. Maintained by hand from pool announcements and observed coinbases, with combined files generated for downstream tools. Useful for attributing blocks to pools; attribution is heuristic, and pools change tags, addresses and infrastructure.

Pool hopping behaviour of Bitcoin miners

A monthly panel of mining pools (blocks, hash rate share, payouts, fees, reward rules), monthly tables of miner addresses hopping between pools, payout pattern statistics, and a block-to-pool attribution table, as CSV files plus one pandas pickle. The authors derived it from the blockchain with BlockSci, published pool tags and a heuristic tracing payouts from coinbase outputs to miners. Useful for studying miner migration; attribution and payout tracing are heuristic.

Solo block record

One row per verified solo-mined block from a small or home miner, with height, pool, hardware, hashrate, difficulty, reward, odds and source links, in CSV and JSON with a Frictionless datapackage. The repository mirrors the dataset SoloLuck.io compiles from miner and press reports, with every block verified on chain. Useful for solo mining odds research; rented-hashrate wins are excluded, and hashrate and derived odds are empty where unreported.

Stratum pool templates

Stratum mining.notify jobs from pool endpoints, one Parquet row per job with observation time, previous block hash, nTime, coinbase, merkle branch count and clean-jobs flag. DataForge collects them from endpoints spanning several pool operators, with a companion table recording collection windows and failures. Useful for studying when pools switch to a new chain tip; the public release is a sample from one vantage point, and full merkle branches are absent.

Development #

Datasets about Bitcoin software development: repositories, issues, pull requests, and the mailing list, chat and forum archives that record it.

Bitcoin Core IRC meeting summaries

One Markdown file per Bitcoin Core development IRC meeting, with YAML front matter and a prose summary of each topic discussed, linking to the raw channel log and minutes. Volunteers wrote them from the #bitcoin-core-dev channel logs and published them in the Bitcoin Core website repository. Useful for reading what the project discussed without parsing IRC logs; summaries exist only for the period volunteers wrote them.

Bitcoin Core PR Review Club meetings

One Markdown file per review club meeting, with YAML front matter naming the reviewed pull request, its author, affected components, the host and the reviewed commit. Written by each meeting's host, who prepares the notes and questions, followed by the full IRC log pasted inline. Useful for studying how Bitcoin Core changes are reviewed and explained; the files are Jekyll posts, so Liquid tags and site-relative links need stripping first.

  • Category: Development
  • Format: Plain text
  • Period: 2019-05-01 – present
  • License: Not specified; meeting log text belongs to its speakers
  • Author: Bitcoin Core PR Review Club hosts and contributors
  • Tags:

Bitcoin Optech topics and newsletters

One Markdown file per Bitcoin technology topic, with YAML front matter giving the title, categories, aliases, primary sources and links to every newsletter that mentioned it, plus the newsletter issues themselves. Bitcoin Optech contributors maintain them in the repository behind the Optech website, against a published topic schema. Useful for mapping protocol topics to the discussions that covered them; the content is editorial prose rather than structured records.

Bitcoin Transcripts

One Markdown file per transcribed talk, podcast episode or developer meeting, with YAML front matter naming the title, speakers, tags, date, the recording it came from and who transcribed it. Volunteers and a machine transcription tool produce them from public recordings, and a review app tracks the editing. Useful for full-text search of what was said about a protocol change; the text is an edited transcription, not verbatim speech.

bitcoin-core/gui GitHub metadata backup

A backup of all issues, pull requests, reviews and comments of the bitcoin-core/gui GitHub repository, where Bitcoin Core GUI changes are developed, one JSON file per issue and pull request as returned by the GitHub API. An automated job refreshes the backup, and a browsable mirror is rebuilt from it. Useful for keeping the development history available independently of GitHub; items GitHub hides or that failed to fetch are listed in state.json.

  • Category: Development
  • Format: JSON
  • Period: 2020-06-16 – present
  • License: Not specified; issue and comment text belongs to its GitHub authors
  • Author: 0xB10C
  • Tags:

bitcoin-core/secp256k1 GitHub metadata backup

A backup of all issues, pull requests, reviews and comments of the bitcoin-core/secp256k1 GitHub repository, the libsecp256k1 elliptic curve library used by Bitcoin Core, one JSON file per issue and pull request as returned by the GitHub API. An automated job refreshes the backup, and a browsable mirror is rebuilt from it. Useful for keeping the development history available independently of GitHub; items GitHub hides or that failed to fetch are listed in state.json.

  • Category: Development
  • Format: JSON
  • Period: 2014-03-06 – present
  • License: Not specified; issue and comment text belongs to its GitHub authors
  • Author: 0xB10C
  • Tags:

bitcoin/bips GitHub metadata backup

A backup of all issues, pull requests, reviews and comments of the bitcoin/bips GitHub repository, the Bitcoin Improvement Proposals, one JSON file per issue and pull request as returned by the GitHub API. An automated job refreshes the backup, and a browsable mirror is rebuilt from it. Useful for keeping the development history available independently of GitHub; items GitHub hides or that failed to fetch are listed in state.json.

  • Category: Development
  • Format: JSON
  • Period: 2013-12-06 – present
  • License: Not specified; issue and comment text belongs to its GitHub authors
  • Author: 0xB10C
  • Tags:

bitcoin/bitcoin GitHub metadata backup

A backup of all issues, pull requests, reviews and comments of the bitcoin/bitcoin GitHub repository, the main Bitcoin Core repository, one JSON file per issue and pull request as returned by the GitHub API. An automated job refreshes the backup, and a browsable mirror is rebuilt from it. Useful for keeping the development history available independently of GitHub; items GitHub hides or that failed to fetch are listed in state.json.

  • Category: Development
  • Format: JSON
  • Period: 2010-12-19 – present
  • License: Not specified; issue and comment text belongs to its GitHub authors
  • Author: 0xB10C
  • Tags:

bitcoindev mailing list archive

A public-inbox mirror of the Bitcoin development mailing list, where one git commit adds one raw email message with its full headers under a path derived from its Message-ID. The mirror covers the list under both its Linux Foundation and Google Groups addresses, and the old pipermail permalinks redirect to it. Useful for offline search of protocol discussion; it is a third-party mirror, not an official archive.

  • Category: Development
  • Format: Plain text
  • Period: 2011-06-12 – present
  • License: Not specified; message text belongs to its email authors
  • Author: gnusha.org
  • Tags:

bitcointalk.org web crawl

WARC files of crawled bitcointalk.org pages, mostly full-thread views of the forum, with gzipped CDX indexes mapping each captured URL to its offset in the WARC files. It comes from a web crawl of the whole forum, uploaded to the Internet Archive's warczone collection. Useful as a bulk capture of early Bitcoin discussion; it is raw crawled HTML in WARC containers, so extracting readable posts takes work.

  • Category: Development
  • Format: Plain text
  • Period: 2017-10-23 – 2017-11-18
  • License: Not specified; the crawled forum posts belong to their authors
  • Author: Internet Archive warczone collection
  • Tags:

en.bitcoin.it wiki dump

A MediaWiki XML export of every page of the en.bitcoin.it wiki with its full revision history, plus title and image lists and the uploaded images, in one 7-Zip archive. WikiTeam collected it through the wiki's MediaWiki API and uploaded it to the Internet Archive. Useful for studying how Bitcoin terminology and documentation were written and revised; it is a one-off capture and no newer dump is published.

  • Category: Development
  • Format: Plain text, Binary
  • Period: until 2014-02-10
  • License: Not specified by the archive item; en.bitcoin.it publishes its pages under CC BY 3.0
  • Author: WikiTeam
  • Tags:

Freenode Bitcoin IRC log archive

A single compressed archive of early IRC logs from the Freenode channels #bitcoin-dev, #bitcoin-core-dev and #bitcoin-otc, one plain text file per channel with a tab-separated timestamp, nickname and message on each line. Bryan Bishop publishes it together with an OpenTimestamps proof of the archive. It covers the years before the daily gnusha.org logs begin; only messages are kept, so joins, parts and other channel events are missing.

  • Category: Development
  • Format: Plain text
  • Period: 2010 – 2018
  • License: Not specified; message text belongs to the channel participants
  • Author: Bryan Bishop
  • Tags:

gnusha.org Bitcoin development IRC logs

Daily plain text logs of the IRC channels used for Bitcoin Core, Lightning and protocol development, one file per channel per day in irssi format with one line per message, join, part or topic change. A logging client sits in the channels and the files are served as plain directory listings. Useful for tracing how a change was discussed; much of the traffic comes from notification bots.

  • Category: Development
  • Format: Plain text
  • Period: 2014-10-19 – present
  • License: Not specified; message text belongs to the channel participants
  • Author: gnusha.org
  • Tags:

Satoshi Nakamoto writings corpus

Structured JSON of Satoshi Nakamoto's emails and forum posts: one record per message with sender, subject, timestamp, body text and a link to its original source, plus thread indexes and selected quotes. The Satoshi Nakamoto Institute collected them from the cryptography and bitcoin-list mailing lists, the P2P Foundation forum and bitcointalk. Useful for citing the original writings; the files also carry other participants' replies, which a field marks apart.

  • Category: Development
  • Format: JSON
  • Period: 2008-10-31 – 2014-03-07
  • License: AGPL-3.0 (repository); the archived writings belong to their authors
  • Author: Satoshi Nakamoto Institute
  • Tags:

Lightning Network #

Datasets about the Lightning Network: channels, gossip, routing.

Early Lightning Network graph snapshots (LNTopology)

Lightning Network graph dumps in the lnd describegraph JSON layout: each node with its public key, alias, announced addresses, reachability flag and coarse geolocation, each channel with its funding outpoint, capacity and the fee policy of both endpoints. The authors pulled them from the lndexplorer graph API for a paper on network topology. These are early snapshots, useful for the period before continuous collections; only a few isolated views exist.

  • Category: Lightning Network
  • Format: JSON
  • Period: 2019-01-03 – 2019-04-18
  • License: Not specified
  • Author: István András Seres, László Gulyás, Dániel A. Nagy, Péter Burcsi (Eötvös Loránd University)
  • Tags:

Geolocated Lightning Network topology snapshots

Lightning Network graph snapshots as NetworkX GML files, where a node carries its public key, alias and city-level geolocation and an edge carries a short channel id, fee policy and HTLC limits, plus a CSV summary table. The authors rebuilt them from the lnresearch gossip archives and geolocated nodes by IP address. Useful for spatial and temporal network analysis; only nodes announcing a public IP address are geocoded.

Lightning channel closings and closing fees

A CSV with one row per closed Lightning channel: short channel id, capacity, closing timestamp, a reason code separating mutual from forced closes, the closing transaction id, the alias of the owning node and the total on-chain fee in satoshis. The author queried the mempool.space API and added the fees of the transactions sweeping the closing outputs. Useful for costing close types; it covers a few large nodes only.

  • Category: Lightning Network
  • Format: CSV
  • Period: 2019-01-20 – 2023-02-20
  • License: MIT (repository); no separate data license stated
  • Author: newabel
  • Tags:

Lightning channel graph snapshot (MPP simulator)

One JSON snapshot of the public Lightning channel graph in lnd describegraph shape: announced nodes with alias, addresses and feature bits, and channels with capacity, funding outpoint and both directional policies. The authors committed it to their multi-path payment simulator, whose parser also reads graphs from the lnresearch topology dataset. It is handy as ready-made simulator input, but a single snapshot without unannounced nodes or channels.

  • Category: Lightning Network
  • Format: JSON
  • Period: 2023-05-16
  • License: MIT (repository); no separate data license stated
  • Author: Charmaine Ndolo and Florian Tschorsch
  • Tags:

Lightning Network cross-layer deanonymization data

A Lightning Network graph snapshot (channels, node public keys, aliases, announced addresses, WHOIS data) with the on-chain funding and settlement transactions of those channels, address-to-entity cluster mappings and the node-to-entity links, as CSV and JSON. The authors collected it from an LND node's describegraph output and a GraphSense instance. Useful for studying how channels relate to on-chain entities; the links rest on heuristics and the graph is a single snapshot.

  • Category: Lightning Network
  • Format: CSV, JSON
  • Period: 2018-01-12 – 2020-09-09
  • License: CC BY 4.0
  • Author: Matteo Romiti, Friedhelm Victor, Pedro Moreno-Sanchez, Peter Sebastian Nordholt, Bernhard Haslhofer, Matteo Maffei (AIT Austrian Institute of Technology, TU Berlin, IMDEA Software Institute, Chainalysis, TU Wien)
  • DOI: 10.5281/zenodo.4482107
  • Tags:

Lightning Network gossip capture (gossip-observer)

Lightning Network gossip messages as sent and received by instrumented LDK nodes, with microsecond timestamps per message, peer and direction, published as monthly Parquet archives. The collectors connect to peers chosen across network communities and export their captures to Zenodo. Useful for studying gossip propagation, fanout and peer sync behaviour; messages sent by collectors are mostly replies to sync requests, not relays.

Lightning Network gossip timings and graph snapshot (lnconv)

A zstd-compressed CSV with one row per observed Lightning gossip message, its type, channel policy fields, origin timestamp and when the collector saw it, plus a zstd-compressed lnd describegraph dump as JSON. The authors connected to many nodes with pyln-proto and took the graph from their own synced LND node. They are the replay input for the authors' gossip simulator, and a one-off recording from a single collector.

Lightning Network listchannels snapshots (ln-probing-simulator)

Core Lightning listchannels dumps as JSON, one record per directed channel with its short channel id, endpoint public keys, capacity, base and proportional fee, CLTV delta and HTLC minimum and maximum. They ship with the probing simulator from a paper on parallel channels and are the graph its experiments run on. Useful as a fixed graph for reproducing probing or routing experiments; channel balances are not gossiped and are absent.

Lightning Network node and channel profiles (Bitcoin Data Labs)

Lightning Network profile tables as Parquet, one row per node or per channel, with capacity, channel counts, fee policy, funding transaction, feature bits and centrality ranks. Bitcoin Data Labs derives them from Lightning gossip and publishes them with a full graph JSON, summary JSON snapshots and a curated file of node entity and role tags. Useful for ranking and classifying routing nodes; the tags and inferred implementation labels are heuristic.

  • Category: Lightning Network
  • Format: Parquet, JSON
  • Period: 2026-09-11 – present
  • License: CC BY 4.0 (data), MIT (scripts)
  • Author: Saurabh Kumar (Bitcoin Data Labs)
  • Tags:

Lightning Network topology gossip datasets (lnresearch)

Lightning Network gossip history: channel_announcement, channel_update and node_announcement messages as raw bzip2-compressed message streams, one file per snapshot. Extracted after deduplication from the gossip_store files of the authors' c-lightning nodes, with checksums and a reader snippet in the README. Useful for reconstructing the network topology and channel policies at any point in time; gossip is lossy by design, so the collection is best effort.