Bitcoin had working oracles long before Ethereum popularized the term—but reconstructing what they actually did is like archaeology, because the blockchain preserves transactions unevenly and often hides the context around them. Before Ethereum made the oracle problem a household term, Bitcoin already had oracles serving as feeds, key-release services, federated signers, and arbiters that carried real value on the main chain.

The study traces their use and the changing evidence of oracle activity from early days through July 2026. It combines a complete census of Counterparty betting with analysis of the full Bitcoin chain and searches for documented service keys in an 854-million-row public-key index.

It also recovers DLC oracle records from an archived explorer and live Nostr relays. In 2011, Hearn proposed making Bitcoin payments conditional on events outside the network, from exchange rates to inheritance. His proposal anticipated agreements whose execution depended on information that the network could not establish on its own.

Over the following years, developers explored this possibility through betting platforms, data feeds, key-release services, and arbitration systems. Research on blockchain oracles has examined the trust, security, and implementation choices involved in connecting contracts to external information.

For Bitcoin, historical research traced a transition from oracle co-signing to designs based on published attestations, particularly discreet log contracts, or DLCs. The scale and distribution of recorded use across these designs remain unevenly documented.

A historical comparison therefore needs to connect evidence of use with the conditions under which that evidence can be observed and interpreted. This study addresses that gap through two research questions. Rather than leaving the identified gap unresolved, this study addresses that gap by organizing its investigation around two clearly stated research questions.

The analysis combines a census of Counterparty betting, analysis of the Bitcoin blockchain from genesis through July 2026, and searches for documented Reality Keys, Orisi, Bitrated, and Oraclize keys in an index of 854 million public-key occurrences. It also covers DLC announcements and attestations recovered from the archived Suredbits explorer and public Nostr relays.

Counterparty activity was concentrated in a small number of services and applications, with modest fee revenues and limited matching of offered value, even though most matched contracts reached settlement. Across the systems examined, measures of activity depend on public keys, information disclosed by transactions, and external records.

This retrospective empirical study examines Bitcoin oracle activity using blockchain transactions, the Counterparty ledger, and public DLC oracle records. Historical documentation and archived websites provide information about service designs and event subjects.

Bitcoin blockchain and Counterparty data were collected in July 2026, with dates expressed in UTC. Reading across the rows, the table distinguishes the observation unit for each component: scripts, public-key occurrences, and transactions for Bitcoin; bets and feeds for Counterparty; and distinct announcement payloads or messages for DLC oracles.

The evidence column also shows a progression from key-matched blockchain records, to Counterparty node and web records, to archived explorer and public Nostr material. Counterparty data were extracted in July 2026 from a local Counterparty Core node connected to a synchronized Bitcoin Core node.

The analysis retrieved valid broadcasts through the REST API and extracted bets, matches, resolutions, expirations, and block records from the node’s SQLite database. Because the ledger records successive states as additional rows, each bet was reconstructed using its first record for placement information and its latest record for final status.

The resulting dataset contains 1,149 bets, including 1,097 protocol-valid submissions, and 397 matches. Matched value was calculated as the cumulative sum of both parties’ contributions across matches, counting funds again whenever they were committed to a subsequent match.

Bet placements and settlements were dated using timestamps from their confirmation and resolution blocks, while each match was dated to the block confirming the second bet in the pair. Bet subjects were classified using broadcast text, feed URLs, archived operator catalogues, and contemporary documentation.

Feed-specific descriptions took precedence over general platform categories, and feeds without sufficient evidence were classified as unrecoverable. The analysis extracted Bitcoin blocks from zero through 958,628, covering January 2009 to July 2026.

Bitcoin Core supplied transactions and previous-output information, which were analyzed using DuckDB and Python. Coverage checks confirmed all 958,629 block heights without gaps or duplicates. The extraction retained bare multisignature, P2SH, P2WSH, Taproot, OP RETURN, and nonstandard outputs, together with scripts revealed by script-path spends.

An index of 854,077,402 public-key occurrences in multisignature inputs and outputs supported exact matching against documented service keys. Each match recorded the transaction identifier, block height, and whether the key appeared in an output or a spending input.

Key occurrences and distinct transactions were counted separately. Scripts were classified by their multisignature requirements, hashlocks, timelocks, and conditional branches. DLC oracle publications were examined through the archived Suredbits explorer and public Nostr relays.

For Suredbits, searches targeted oracle, event, announcement, and attestation pages on the main and test domains. The analysis extracted oracle names, event descriptions, maturity dates, outcomes, and embedded payloads. Bitrated arbitration was included as an application related to oracles, while services that co-sign transactions solely for wallet security and custodial gambling platforms fell outside the study’s scope.

OpenBazaar could not be examined through key matching because moderator keys were not recovered. Public-key matching identifies documented service keys in indexed Bitcoin scripts. Counterparty records provide measures of betting activity, matched value, and settlement, while DLC records document oracle announcements and attestations.

The first bet recorded in the Counterparty ledger was placed on 2 February 2014, as a one-XCP wager on a feed labeled Super Bowl Bet Feed: one equals Seattle Won. Activity was concentrated in 2014, which accounted for 1,105 of the 1,149 recorded bet transactions, or 96.2 percent.

June alone contained 465 bets, or 40.5 percent of the full dataset, overlapping with the FIFA World Cup held from 12 June to 13 July. Only 44 bets were recorded after 2014, and no later submissions appear in the ledger examined through July 2026. The chart’s central cluster falls around June and October twenty fourteen, while the later timeline is nearly flat, with only small isolated bars in twenty fifteen through twenty seventeen.

The tallest June bar reaches just under five hundred bets, whereas the October peak rises a little above two hundred, making the two main surges visibly distinct. Across the observation period, sports accounted for 90.5 percent of matched XCP value, followed by price feeds at 8.3 percent and gambling feeds at 1.2 percent.

Most submitted bets used the binary Equal or NotEqual types, accounting for 1,033 of the 1,149 records, or 89.9 percent. The remaining 116 were contracts for difference, of which 86 were protocol-valid. The invalid submissions included 29 rejected with the status invalid: CFDs temporarily disabled.

In the horizontal bars, the sports category stretches nearly to the ninety-five percent gridline, while the price-feed bar ends just below ten percent. The gambling, other or unrecoverable, and test bars are compressed against the zero end of the axis, making their visual contribution much smaller.

Across the 98 feeds with matched activity, the Herfindahl–Hirschman Index, calculated from squared percentage shares of matched value, was 711. When feeds were grouped by operator, the Herfindahl–Hirschman Index was 7,787.

The 82 matched feeds associated with xbet.io represented 87.6 percent of total matched value, while xcpfeeds.info accounted for 10.6 percent. Together, these two services accounted for 98.2 percent of matched value.

In panel a, the first bar nearly spans the full horizontal scale, while every remaining bar is much shorter, with the next largest reaching only a little beyond ten percent. In panel b, the blue Lorenz curve stays close to the bottom before bending sharply upward near the right edge, well below the dashed equal-shares line.

Across the full betting history, users offered 7,158 XCP, but only 2,625 XCP were matched, equivalent to 36.7 percent of the total offered. Of the 1,097 valid bets, 588, or 54 percent, ended with an expired status, 38 percent were filled, and 8 percent were canceled; none remained open.

Once bets were matched, most contracts reached settlement: 387 of the 397 matches, or 97.5 percent, were resolved. Nine matches expired, returning approximately 19.8 XCP through the protocol’s refund mechanism. Across blocks zero through 958,628, the chain analysis identified 450,015,065 spends that revealed scripts within the defined categories.

Of these, 316,060,091 revealed multisignature scripts. This large population provided context for searching for oracle-related activity, although the same script structures were shared by many applications. The analysis flagged 842,241 bare multisignature outputs containing at least one public-key value that could not be decoded as a valid curve point.

Matching documented public keys against the chain index located transactions associated with Orisi, Oraclize, and Bitrated, while none of the recovered Reality Keys matched. For Orisi, seven of the ten recovered keys appeared in 17 mainnet transactions between 12 June and 14 October 2014.

Oraclize’s documented service key appeared in 106 distinct transactions dated 2015 to 2016, all containing 2-of-3 scripts. For Bitrated, 11 of the 11,261 recovered public keys appeared in 56 distinct transactions, and the overall match rate was 0.10 percent. The clearest comparison is in the “Observed script types” column: Orisi spans four script forms, whereas Oraclize and Bitrated each show only two-of-three scripts.

The Reality Keys row is qualitatively distinct, pairing “No matches found” with “Not applicable” rather than a script type or transaction period. Taproot key-path spends disclose a signature without revealing a spending script.

Within the defined categories, their share rose from 2.6 percent in 2022 to 67.4 percent in 2024, before declining to 52.8 percent in 2025 and 50.8 percent in the partial 2026 record. The series was sensitive to the treatment of inscription-related scripts.

Between 2023 and the 2026 cutoff, 92 to 98 percent of Taproot script-path spends contained the Ordinals marker. The chart’s two curves overlap through 2022, then separate sharply: the inscription-corrected series sits above the raw series in each subsequent year.

The widest visible gap occurs in 2024, at roughly fifteen percentage points, before narrowing somewhat in 2025 and 2026. The analysis covered 3,270 records extracted from archived announcement and event pages of the Suredbits explorer, with capture dates between 2019 and 2023.

The production-domain records yielded 911 distinct announcement payloads. Of these, 673, or 73.9 percent, were associated with attestation records in the recovered pages. Recorded event maturities were concentrated in 2021, with 282 announcements, and 2022, with 446.

Within the recovered record, price and finance accounted for 649 announcements, or 71.2 percent, while sport accounted for 170, or 18.7 percent. Much of the concentration in price and finance came from the recurring Deribit-BTC reference series, which contributed 597 announcements, or 65.5 percent of the corpus.

Excluding this series left 314 announcements, with sport accounting for 54.1 percent and price and finance for 16.6 percent, reversing their ranking. The publication mix was therefore strongly influenced by one recurring series.

In the sport group, the hatched bar rises to fifty-four point one percent, while the solid bar is labeled eighteen point seven percent. The price-and-finance pair shows the opposite visual pattern: the solid bar reaches seventy-one point two percent, whereas the hatched bar is sixteen point six percent.

Sports account for 83.5 percent of Counterparty bets with identified feeds and 90.5 percent of matched XCP value. Price and finance represent 71.2 percent of Suredbits announcements, largely due to the recurring Deribit-BTC series. Excluding this series makes sport the largest category, accounting for 54.1 percent of the remaining announcements.

In the Nostr collection, tests and demonstrations predominate, followed by delivery and return labels published under a single key. The study reconstructs Bitcoin oracle activity across early services and more recent DLC publications. The Counterparty census documents 1,149 bet submissions between 2014 and 2017, with activity concentrated in 2014 and matched value dominated by two services.

Most matched contracts reached settlement, although much of the offered value remained unmatched and fee revenues were modest. The findings show how contract design shapes the evidence available for measuring use. Interpreting this evidence requires distinguishing publication activity from the contracts that consume it and considering how that relationship changes across implementations.

The big lesson is that measuring oracle use requires more than counting transactions: protocol design, public keys, archived descriptions, and privacy all determine what historical evidence survives and what remains invisible.