OutcomeDepth mark OutcomeDepth Historical data
Historical data infrastructure · crypto + weather

Historical Polymarket data infrastructure for analysis, simulation, and integration.

A normalized, round-level archive for trading firms, analytics platforms, and data teams. Build backtests, models, and internal data services from documented Arrow files containing probabilities, visible depth, timing, outcomes, and quality evidence.

Bulk-readyDocumented Arrow archives
Integration-readyPython, Rust, and internal APIs
Audit-readyTiming and quality evidence
53,432 standardized files 7 crypto assets Weather markets Capture growing daily
Inside one BTC round file T−02:41
Five-minute market state · synthetic format preview
Observed up probability 0.64 ▲ +0.012
00:0002:3005:00
Bids
0.63180
0.62120
0.6164
0.6042
0.5927
Spread
0.02
Asks
1460.65
1010.66
580.67
420.68
250.69
300normalized seconds 46live observations UP TO 20levels per side UTCevent timestamps
observed_at seconds_remaining up_probability down_probability best_bid best_ask book_depth book_age_ms observed/resampled outcome quality
53,432
Standardized round files
22,093
Detailed BTC rounds
7+1
Crypto assets + NYC weather
Daily
Capture continues
Audited snapshot · 10 Aug 2026. Coverage varies by collection, date, field and capture quality.
BTC 22,093 detailed rounds · 97.52% of slots ◆ ETH 6,493 ◆ SOL 5,377 ◆ XRP 5,378 ◆ BNB 4,909 ◆ DOGE 4,578 ◆ HYPE 4,604 ◆ Weather · NYC high & low · since 10 Aug ◆ Apache Arrow · round-level files · quality flags ◆ audited snapshot 10 Aug 2026 ◆ BTC 22,093 detailed rounds · 97.52% of slots ◆ ETH 6,493 ◆ SOL 5,377 ◆ XRP 5,378 ◆ BNB 4,909 ◆ DOGE 4,578 ◆ HYPE 4,604 ◆ Weather · NYC high & low · since 10 Aug ◆ Apache Arrow · round-level files · quality flags ◆ audited snapshot 10 Aug 2026 ◆
Why it matters

A correct prediction is not the same as a tradable one.

A probability chart tells you what a market displayed. A timestamped market-state archive lets you ask whether the price was fresh, whether size was actually available, and whether an apparent edge survives spread, depth and missing data.

Price-only view Assumption-heavy
Model output · Up
64%
Displayed market price: 64¢
SpreadUnknown
Available sizeUnknown
Observation ageUnknown
Data gapsHidden
Backtest decision
Buy at 64¢

Assumes the displayed probability was an executable price.

Market-state view Execution-aware
Model output64% → Best ask65¢ the displayed price was not the executable price
Ask ladderCumulative size
65¢18 available
66¢31 available
67¢51 available
68¢78 available
69¢112 available
70¢156 available
Spread2¢ At best ask18 Target size100 Source age1.4s
300 secondsnormalized series Live snapshotsobserved separately Outcome + qualityround context
Data-aware decision
Size down · wait · skip

Tests the signal against the market that could actually be traded.

Synthetic illustration · not a trading result, profit claim or archived market record

Avoid false edges
Reject impossible fills.

A signal can look profitable only because a backtest assumes unlimited size at a displayed price.

Test better features
See more than probability.

Explore whether spread, depth, imbalance, timing and data quality improve calibration or signal selection.

Control downside
Know when not to trade.

Separate model failure from stale observations, thin liquidity, missing evidence and execution constraints.

Packages & pricing

Choose the coverage.
Then choose the depth.

Core is a compact, analysis-ready subset. Full Depth preserves the maximum permitted market-state detail. Every annual option includes the approved back catalogue plus 12 months of releases.

One asset
Bitcoin 5-minute archive
Core€550 /yr
Full depth€800 /yr

Begin with BTC only — compact analytical schema or the complete permitted BTC structure.

Get BTC access
All crypto
Multi-asset crypto archive
Core€1,050 /yr
Full depth€1,400 /yr

BTC, ETH, SOL, XRP, BNB, DOGE and HYPE, with documented asset-specific coverage dates.

Get 7-asset access
Weather · 1 location
Single-location weather
Core€550 /yr
Full depth€800 /yr

One documented city or reference-station collection with its daily high and low markets.

Get 1-location access
Weather · 3 locations
Three-location weather
Core€850 /yr
Full depth€1,200 /yr

Up to three documented locations, daily high and low markets organized consistently.

Get 3-location access
Best value
Crypto + weather
OutcomeDepth Complete
Core · 3 loc€1,350 /yr
Full · all loc€1,999 /yr

All supported crypto assets plus weather — every approved location on Full Depth.

Get Complete access
Individual internal use No redistribution or resale Files retained under licence Updates end without renewal

Proposed individual pricing · availability subject to written rights, eligibility and final terms

Know exactly what the upgrade adds.

Core is a defined subset — not an undocumented reduction. Full Depth adds the detailed observations and nested market structure needed for higher-fidelity replay and execution-aware simulation.

Arrow filesgrouped by collection and date
Manifestmarkets, paths, status, coverage
Schemafields, types, nesting, definitions
Quality reportdensity, missingness, limitations
Checksumsintegrity hashes per file
Release notesversions, corrections, changes
Data componentCoreFull depthOther providers*
Market identity and timing✓✓Common
Market state paired with final outcome✓✓Not standard
Probabilities and top of book✓✓Common
Per-round capture-quality flags✓✓Not standard
Observed vs normalized layer labels✓✓Not standard
Detailed observation snapshots—✓Plan-dependent
Visible order-book ladders—✓Plan-dependent
Observation-age evidence inside each round—✓Not standard
Self-contained nested Arrow file per round✓✓Not standard
Unified crypto + weather option✓✓Not standard

* General category comparison based on standard market-data products reviewed in August 2026. “Not standard” means the capability is not ordinarily presented as a default product feature—not that no vendor or custom plan can provide it. Coverage still depends on the selected OutcomeDepth package.

1 Share your requirementsMarkets, date range, package scope, and preferred delivery workflow.
2 Confirm the data fitAvailable coverage, schema, integration options, and relevant limitations.
3 Agree access detailsSelect Core or Full Depth and confirm commercial and delivery terms.
Coverage

A growing archive, described honestly.

Coverage is not uniform: crypto and weather have different market structures, capture histories and constraints. Those differences are documented per collection, date and field — never flattened into one completeness number.

Finalized five-minute round files by asset 24 MAY → 10 AUG 2026
BTC24 May22,09397.52%
ETH18 Jul6,49398.02%
SOL22 Jul5,37799.37%
XRP22 Jul5,37899.39%
BNB24 Jul4,90999.09%
DOGE25 Jul4,57898.49%
HYPE25 Jul4,60499.05%
BTC file availability · 79 days 97.52% · 22,093 / 22,654 slots
24M31M07J14J21J28J05J12J19J26J02A09A*

Weekly file availability, % of five-minute slots. * Final bar is a partial two-day window. File availability does not imply uniform tick density inside every file.

Weather · NYC New collection

Daily highest- and lowest-temperature markets with forecast context, station evidence and settlement notes. Capture since 10 Aug 2026.

83.4°Fforecast high 79.2°Fobserved so far 4.0 hto market close 6market observations 2forecast snapshots 42 slatest source age
≤ 79°F0.08
80–81°F0.19
82–83°F0.31
84–85°F0.25
≥ 86°F0.17
bucket bid / ask forecast issued_at station evidence settlement context quality summary

Illustrative distribution · not an archived record

Why the qualifiers matter. A timestamp in the inventory does not guarantee a complete event tape, full-depth book, or an independently observed value for every second. Coverage is published by collection, date, field, source type and observation quality before access is offered.

Arrow file anatomy

One round.
Nested evidence.

A crypto file is one self-contained round record: identity, timing, outcome context, live observations, a normalized second-by-second series, deep book samples, and a quality summary.

crypto-5m-synthetic.arrow
├─ TOP LEVELidentity · timing · outcome ├─ LIVE TICKS46 observed snapshots ├─ SECONDS300 normalized elements ├─ BOOKS20 bid + 20 ask levels └─ QUALITYobserved · resampled · age
1top-level row 20top-level fields Nestedlists + structs
Decoded PyArrow object · synthetic excerpt Format-faithful
[
  {
    "synthetic": true,
    "schema_version": "synthetic-crypto-2",
    "collection": "crypto_5m",
    "market_id": "synthetic-crypto-round-001",
    "market_slug": "synthetic-asset-up-or-down-5m",
    "asset": "SYNTHETIC_CRYPTO",
    "interval_seconds": 300,
    "market_start": "2031-01-01T12:00:00Z",
    "market_end": "2031-01-01T12:05:00Z",
    "resolution_status": "synthetically_resolved",
    "winning_side": "up",
    "observed_snapshot_count": 46,
    "normalized_second_count": 300,
    "live_tick_snapshots": [
      {
        "second": 0,
        "observed_at": "2031-01-01T12:00:00Z",
        "seconds_remaining": 299,
        "up_probability": 0.48,
        "down_probability": 0.52,
        "up_bid": 0.465, "up_ask": 0.495,
        "book_age_ms": 70,
        "is_observed": true,
        "is_resampled": false,
        "quality": "observed"
      },
      … 45 additional captured snapshots
    ],
    "normalized_seconds": [
      … 300 observed or explicitly resampled elements
    ],
    "order_book_samples": [
      {
        "token_side": "up",
        "bids": [{"price": 0.63, "size": 180.0}, … 19 more],
        "asks": [{"price": 0.65, "size": 18.0}, … 19 more]
      }
    ],
    "quality_summary": {
      "observed_seconds": 46,
      "resampled_seconds": 254,
      "max_book_age_ms": 379,
      "coverage_percent": 15.33
    }
  }
]
Synthetic structure preview. Identifiers, timestamps, prices, sizes, outcomes, and measurements are fabricated.
Inspect before enquiring

Open the exact format first. Free.

Two synthetic Arrow files show the Full Depth structure — entirely fabricated identifiers, prices and outcomes. Load them in Python, Rust, SQL or any Arrow-compatible stack; no proprietary viewer.

Synthetic format samples only · not a free data tier · not suitable for prediction, backtesting or trading

crypto-5m-synthetic.arrowOne full five-minute round · 46 live snapshots · 300-second series · 20-level book · quality Download
weather-nyc-synthetic.arrowOne NYC temperature record · buckets · forecasts · settlement · quality Download
TOP LEVEL · identity · timing · outcome LIVE TICKS · books · probabilities SECONDS · 300 normalized QUALITY · observed vs resampled
Method & limits

Reliable data starts with what was observed.

Captured values, resampled values, derived features and final outcomes are kept separate. The goal is not to make every row look complete — it is to make the evidence behind each row clear.

01Raw evidence
Observed events

Timestamped snapshots and event messages actually received.

02Analysis-ready
Normalized state

Consistent field names, market sides and time alignment.

03Not new observations
Fixed-time grid

Resampled or forward-filled rows, explicitly identified.

04Revision-aware
Outcome context

Resolution status and settlement evidence attached later.

Contact us

Discuss your data requirements.

1Coverage and availability for the required markets and dates
2Package, schema, and integration fit for your workflow
3Commercial details, relevant limitations, and next steps
DiscordView the community → TelegramStart an enquiry → EmailSend a data question →

Early-access enquiry · no payment requested · availability subject to rights and review

OutcomeDepth mark OutcomeDepth

Independent data archive. Not affiliated with or endorsed by Polymarket or any exchange, oracle, weather service or data provider. Historical data · no advice · no performance claims.

Access & legal notice Back to top ↑