Data infrastructureLiveData & Developer Tools

OpenMarket

Reproduce Polymarket–Binance microstructure research from a frozen public corpus.

Polymarket support
Polymarket native
Pricing model
Free · Scope and terms apply
Access requirement
Open access
Wallet required
No
Trade execution
None
Last editorial review
Nov 6, 2026

Which OpenMarket does this profile cover?

OpenMarket pairs millisecond-level Polymarket BTC binary-market events with Binance BTC/USDT data and publishes a versioned Parquet corpus, Rust research pipeline and walk-forward baselines.

Official product location: github.com/gregyoung14/openmarket

Scope and limits: Collection is frozen and covers a bounded 2026 BTC binary-market period, with missing days, reconnect risk and a remaining constant clock-offset ambiguity. Top-of-book simulations may overstate fills. The published model slightly underperformed the Polymarket mid-price baseline and had negative simulated economics; the full archive and pipeline were not rerun by PredAtlas.

Inspect the dated sources and access requirements for this product before using it.

Read a small OpenMarket Polymarket dataset sample

This check reads one frozen demo file: lag_pairs_ms.parquet, 25,171 bytes and 1,769 paired observations. It checks actual field types, timestamp coverage and the lead–lag formula before you attempt the full research archive.

Executed by PredAtlas on , using Python 3.12.2 and PyArrow 19.0.1 on macOS arm64. This is a bounded data check, not a rerun of the paper or a full product review.

Download only the demo file

Use a new local folder with Python 3.12 and curl installed. The dataset URL pins a specific revision. The download is capped at 1 MiB; the checker verifies the SHA-256 below and refuses more than 10,000 rows. It makes no network or trading calls.

python3 -m venv .venv
.venv/bin/python -m pip install pyarrow==19.0.1
curl --fail --location --max-time 60 --max-filesize 1048576 \
  'https://huggingface.co/datasets/gregyoung14/openmarket-btc-polymarket/resolve/74502466d1a7cef56395bfd8d0b465fbebc849cf/lag_pairs_ms.parquet' -o lag_pairs_ms.parquet
curl --fail --max-time 30 --max-filesize 65536 \
  'https://predatlas.com/examples/openmarket-sample-check.py' -o openmarket-sample-check.py
.venv/bin/python openmarket-sample-check.py lag_pairs_ms.parquet

Inspect the complete checker before running it. Expected sample SHA-256: dd5a7b4c1f0ab82fa355a86325459359399f21029ee3a7eb578c0e994e4ad2d4.

Observed output

rows=1769; bytes=25171; missing_required=0
paired_at_utc=2026-05-14T14:58:42.554Z..2026-05-14T14:59:28.138Z
binance_source_utc=2026-05-14T14:58:27.479Z..2026-05-14T14:59:17.062Z; polymarket_source_utc=2026-05-14T14:58:27.235Z..2026-05-14T14:59:16.961Z
lead_lag_ms=-613..508; formula_mismatches=0
quality_flag:string={'medium': 571, 'tight': 1078, 'wide': 120}; side_label=['DOWN', 'UP']

The pairing timestamps span 45.584 seconds on May 14, 2026. Source-event timestamps precede these pairing timestamps; they are different clocks and must not be treated as interchangeable event times. All 1,769 rows satisfy lead_lag_ms = polymarket_source_ts_ms - binance_source_ts_ms.

Two schema details to check before filtering

In this file, quality_flag is a string with tight, medium and wide values. The reference README describes an integer flag. Likewise, actual side_label values are uppercase UP and DOWN, while the reference shows lowercase labels. Inspect the file schema and values instead of assuming a numeric quality filter or lowercase match.

The file contains integer IDs and millisecond timestamps; text market, side, quality and date fields; and floating-point prices. This check does not validate foreign keys against the other sample tables, complete order books, clock synchronization, executable fills or trading returns. Positive lead–lag means a later Polymarket source timestamp, not a demonstrated tradable edge. This short demo is not representative of the full archive’s coverage or gaps.

Source and attribution: Gregory Young, OpenMarket BTC Polymarket, pinned dataset revision, root v0.1-sample demo. Dataset card declares Apache-2.0; see the source license and v0.5.2 dataset reference. Only derived checks are shown here; download the original file from its publisher.

01 · Overview

What is OpenMarket?

OpenMarket pairs millisecond-level Polymarket BTC binary-market events with Binance BTC/USDT data and publishes a versioned Parquet corpus, Rust research pipeline and walk-forward baselines. Best for: Researchers testing short-horizon crypto-market microstructure, timestamp alignment, order-book response, feature engineering and reproducible backtests. Less suitable for: Users seeking current all-market coverage, a hosted live signal, maintained production trading software or evidence of deployable forecasting alpha. Key limitations: Collection is frozen and covers a bounded 2026 BTC binary-market period, with missing days, reconnect risk and a remaining constant clock-offset ambiguity. Top-of-book simulations may overstate fills. The published model slightly underperformed the Polymarket mid-price baseline and had negative simulated economics; the full archive and pipeline were not rerun by PredAtlas.

Historical DataMarket Analytics

02 · Best for

Who is OpenMarket best for?

Best for

Researchers testing short-horizon crypto-market microstructure, timestamp alignment, order-book response, feature engineering and reproducible backtests.

Not ideal for

Users seeking current all-market coverage, a hosted live signal, maintained production trading software or evidence of deployable forecasting alpha.

Limitations

Collection is frozen and covers a bounded 2026 BTC binary-market period, with missing days, reconnect risk and a remaining constant clock-offset ambiguity. Top-of-book simulations may overstate fills. The published model slightly underperformed the Polymarket mid-price baseline and had negative simulated economics; the full archive and pipeline were not rerun by PredAtlas.

03 · Access & execution

How do access and execution work?

Entity typeData infrastructure
Polymarket supportPolymarket native
Pricing modelFree · Scope and terms apply
Free planYes
Access requirementOpen access
Wallet requiredNo
Trade executionNone
Product statusLive
Polymarket

Scope-aware product facts

What to look for

Surfaces

Public-Versioned-DatasetWeb · Supported Evidence
Rust-Research-WorkspaceDesktop · Supported Evidence
Research-PaperWeb · Supported Evidence

Access modes

Source-And-DatasetOpen access · wallet No · Not required · Not applicable Evidence

Execution modes

Published-Data-And-Methods-ReleaseNone · Not supported · Not applicable Evidence

Pricing offers

Source-CodeFree-Content · $0 · Supported · Apache-2.0 licensed source; local compute, memory and storage are user-supplied. Evidence
Published-DatasetFree plan · $0 · Supported · The full unified Parquet corpus is approximately 8.7 GiB; download, storage and compute costs remain with the user. Evidence

Data capabilities

Synchronized-Cross-Venue-Corpus · Historical-DataSupported · Versioned Parquet Splits Pair Polymarket BTC Binary-Market Ticks With Binance BTC/USDT Events Using Both Source And Collector Timestamps; Collection Is Frozen. Evidence
Order-Book-Replay · Historical-DataSupported · The Unified Split Includes Deduplicated Polymarket Ticks, Binance Ticks And Trades, Market Metadata And Explicit Millisecond Lead-Lag Pairs For A Bounded 2026 Observation Period. Evidence
Reproducible-Research-Pipeline · Market-AnalyticsSupported · Published Collectors, Exporters, Validation Scripts, Walk-Forward Training And Backtesting Components Are Tied To Source, Dataset, Model And Configuration Versions. Evidence
Published-Null-Baseline · Market-AnalyticsSupported · The Paper Reports That Its Walk-Forward Model Slightly Underperformed The Naive Polymarket Mid-Price Prior And Produced Negative Simulated Economics Under Stated Assumptions. Evidence

04 · Data provenance

Data provenance

Current coverage: CompleteA transparent record of which official sources were checked and when.

Data originMixed sources
PnL methodologyNot applicable
Fees includedNot applicable
Rewards includedNot applicable
Open positions includedNot applicable
Order-book historyRecorded
Update cadenceNot applicable
Review provenance evidence

05 · Public footprint

Public footprint

Current public signals from official platform sources, shown with their observation dates.

Product activity

A transparent record of which official sources were checked and when.

Product statusLive
Tracked by PredAtlas sinceSep 8, 2026
Last editorial reviewNov 6, 2026
Latest public updateJul 31, 2026
Official documentationYes
Public repositoryYes

06 · Sources & review

How was OpenMarket reviewed?

Documentation strength: CompleteA transparent record of which official sources were checked and when.

Sources reviewed · Nov 6, 2026

Official product sources were checked for documented identity, scope, access, pricing and limitations. This is not a security audit, performance test or endorsement.