Settled funding rates
Binance USD-M · native settlement intervals
120,684 retained rows24 markets · 2021-12-01 → 2026-08-31
Reconstruct carrying costs and compare funding regimes. These are settled observations, not projected rates.
THE COINNUDGE DATA CATALOGUE
Crypto data coverage. One catalogue, clear access.
The single catalogue for CoinNudge data: compare ten purchasable historical research packs, their actual fields, symbols and dates. Free current Research and other internal inventories are labelled separately. Choose a fixed edition, not a monthly API tier.
Which crypto datasets can you actually use?
Use the catalogue below to separate publicly accessible observations, published datasets included in the named data packs, and other internal sources requiring separate delivery review. A stored row count is not a purchase entitlement or a guarantee of uninterrupted history.
Frozen input inventory · Candidate batch B20260928T024406-038815 sealed 2026-09-28. New source updates belong to later batches.
Open interest history
Binance USD-M · 5-minute observations
10,605,126 retained rows24 markets · 2021-12-01 → 2026-09-26
Study leverage build-up and unwinds. Contract quantity and notional are different measurements.
Historical market candles
Binance Spot · interval-specific coverage
44,344,972 retained rows391 markets · 2021-08-01 → 2026-09-28
Build return, volume and volatility features. Earliest dates and market counts differ by interval.
Public Research datasets
67 current datasets cover market context, derivatives, flows, events and research results. Each dataset has its own source, update cadence and coverage boundary.
Registered free API key: up to 60 requests/minute. Current JSON/CSV access does not unlock paid history.
Read free Research API methods →Market Events v1
Recorded public events, scheduled comparison observations, available context and later price outcomes. Existing legacy plans include this event dataset; their additional historical feature coverage differs.
BTC, ETH, SOL, BNB, XRP and DOGE · Binance USDT Spot. Published history only, not a multi-year candle archive.
Inspect new data pack samples →10 research data packs. Five complementary categories.
Choose by the question you want to test. These are processed, column-scoped historical products—not ten unrelated raw datasets. Related packs share source observations and, where supplied, forward-outcome labels.
Funding cost history
- What data is this?
- Historical settled funding for 24 supported perpetual markets, normalized to an hourly rate with observed 1-, 7- and 30-day sums and completeness flags.
- Why is it useful?
- Calculate and compare the funding component of holding a perpetual position across markets and periods without rebuilding every settlement window yourself.
- Research question
- Which contracts and periods carried the largest observed funding cost, and which windows are incomplete?
- What it cannot prove
- Funding is only one carrying-cost component. These rows do not include your executed fills, trading fees, borrowing cost or net strategy profit.
| Published feature range (UTC) | 2021-12-01 → 2026-08-31. Endpoints do not prove continuous coverage. |
|---|---|
| Included symbols | AAVEUSDT · ADAUSDT · ARBUSDT · ATOMUSDT · AVAXUSDT · BCHUSDT · BNBUSDT · BTCUSDT · DOGEUSDT · DOTUSDT · ETCUSDT · ETHUSDT · FILUSDT · INJUSDT · LINKUSDT · LTCUSDT · NEARUSDT · OPUSDT · SOLUSDT · SUIUSDT · TRXUSDT · UNIUSDT · XLMUSDT · XRPUSDT |
| Rows | 120,684 features · 0 separate forward labels. Repeated horizons are not independent observations. |
| Bulk files | 7.9 MiB Parquet ZIP, plus an Excel/WPS-ready CSV ZIP split at 500,000 data rows per file. Both include coverage and opening instructions. |
Delivered feature fields: record_key, symbol, observed_at, ingested_at, feature_time, historical_retrospective, method_version, hourly_equivalent_pct, funding_interval_hours, settled_at, gap_before, observed_sum_pct_1d, observed_sum_pct_7d, observed_sum_pct_30d, settlement_n_1d, settlement_n_7d, settlement_n_30d, window_complete_1d, window_complete_7d, window_complete_30d
Inspect intervals, per-partition dates and rows in the bundled coverage report. Hourly/daily price features and native settlements must not be mixed without alignment.
Funding extremes & persistence
- What data is this?
- The same supported funding-settlement history transformed into trailing 7-, 30- and 90-day percentiles, sign changes and positive/negative streaks.
- Why is it useful?
- Identify when funding was unusual relative to that contract's own prior history, rather than comparing an isolated rate with an arbitrary fixed threshold.
- Research question
- Was a high funding reading genuinely extreme for this coin, and how long had its sign persisted?
- What it cannot prove
- This is not a future-return label or proof that crowded positions will reverse. Early or gapped percentile windows may be unavailable.
| Published feature range (UTC) | 2021-12-01 → 2026-08-31. Endpoints do not prove continuous coverage. |
|---|---|
| Included symbols | AAVEUSDT · ADAUSDT · ARBUSDT · ATOMUSDT · AVAXUSDT · BCHUSDT · BNBUSDT · BTCUSDT · DOGEUSDT · DOTUSDT · ETCUSDT · ETHUSDT · FILUSDT · INJUSDT · LINKUSDT · LTCUSDT · NEARUSDT · OPUSDT · SOLUSDT · SUIUSDT · TRXUSDT · UNIUSDT · XLMUSDT · XRPUSDT |
| Rows | 120,684 features · 0 separate forward labels. Repeated horizons are not independent observations. |
| Bulk files | 5.7 MiB Parquet ZIP, plus an Excel/WPS-ready CSV ZIP split at 500,000 data rows per file. Both include coverage and opening instructions. |
Delivered feature fields: record_key, symbol, observed_at, ingested_at, feature_time, historical_retrospective, method_version, hourly_equivalent_pct, settled_at, funding_interval_hours, gap_before, percentile_7d, percentile_30d, percentile_90d, prior_n_7d, prior_n_30d, prior_n_90d, sign_changed, sign_streak_count, sign_streak_hours, streak_left_censored
Inspect intervals, per-partition dates and rows in the bundled coverage report. Hourly/daily price features and native settlements must not be mixed without alignment.
Open interest changes
- What data is this?
- Historical open-interest observations for 24 supported perpetual markets, with 1-hour, 4-hour, 24-hour and 7-day changes in contract quantity and price-sensitive notional.
- Why is it useful?
- Separate real changes in outstanding contract quantity from notional changes caused partly by price, while retaining missing-interval flags.
- Research question
- Did open interest actually expand as contracts, or did its dollar value mostly move with price?
- What it cannot prove
- Open interest does not reveal whether new positions belong to longs or shorts, and gaps cannot be treated as zero change.
| Published feature range (UTC) | 2021-12-01 → 2026-09-26. Endpoints do not prove continuous coverage. |
|---|---|
| Included symbols | AAVEUSDT · ADAUSDT · ARBUSDT · ATOMUSDT · AVAXUSDT · BCHUSDT · BNBUSDT · BTCUSDT · DOGEUSDT · DOTUSDT · ETCUSDT · ETHUSDT · FILUSDT · INJUSDT · LINKUSDT · LTCUSDT · NEARUSDT · OPUSDT · SOLUSDT · SUIUSDT · TRXUSDT · UNIUSDT · XLMUSDT · XRPUSDT |
| Rows | 10,605,126 features · 0 separate forward labels. Repeated horizons are not independent observations. |
| Bulk files | 906.7 MiB Parquet ZIP, plus an Excel/WPS-ready CSV ZIP split at 500,000 data rows per file. Both include coverage and opening instructions. |
Delivered feature fields: record_key, symbol, venue, event_time, observed_at, ingested_at, feature_time, source_available_at, historical_retrospective, method_version, valid_current, previous_gap, oi_quantity_change_1h_pct, oi_quantity_change_4h_pct, oi_quantity_change_24h_pct, oi_quantity_change_168h_pct, oi_notional_change_1h_pct, oi_notional_change_4h_pct, oi_notional_change_24h_pct, oi_notional_change_168h_pct
Inspect intervals, per-partition dates and rows in the bundled coverage report. Hourly/daily price features and native settlements must not be mixed without alignment.
Price–OI regimes
- What data is this?
- Time-aligned spot-price returns and perpetual open-interest quantity changes, classified into price/OI quadrants for the supported markets.
- Why is it useful?
- Compare rising-price/rising-OI periods with falling-price/falling-OI or other combinations using one documented classification instead of manually joining two histories.
- Research question
- Was a move accompanied by expanding leverage, falling open interest or conflicting evidence?
- What it cannot prove
- A quadrant is descriptive. It does not identify long/short ownership, liquidations or a tradeable signal by itself.
| Published feature range (UTC) | 2021-12-01 → 2026-09-26. Endpoints do not prove continuous coverage. |
|---|---|
| Included symbols | AAVEUSDT · ADAUSDT · ARBUSDT · ATOMUSDT · AVAXUSDT · BCHUSDT · BNBUSDT · BTCUSDT · DOGEUSDT · DOTUSDT · ETCUSDT · ETHUSDT · FILUSDT · INJUSDT · LINKUSDT · LTCUSDT · NEARUSDT · OPUSDT · SOLUSDT · SUIUSDT · TRXUSDT · UNIUSDT · XLMUSDT · XRPUSDT |
| Rows | 10,605,126 features · 0 separate forward labels. Repeated horizons are not independent observations. |
| Bulk files | 245.2 MiB Parquet ZIP, plus an Excel/WPS-ready CSV ZIP split at 500,000 data rows per file. Both include coverage and opening instructions. |
Delivered feature fields: record_key, symbol, venue, event_time, observed_at, ingested_at, feature_time, source_available_at, historical_retrospective, method_version, valid_current, previous_gap, spot_return_1h_pct, oi_quantity_change_1h_pct, oi_price_quadrant, spot_reference_venue
Inspect intervals, per-partition dates and rows in the bundled coverage report. Hourly/daily price features and native settlements must not be mixed without alignment.
Unusual trading volume
- What data is this?
- Spot quote-volume observations expressed relative to a trailing baseline, with price context, eligible non-event rows and separate forward-outcome files where available.
- Why is it useful?
- Screen genuinely unusual volume across different-size coins and compare later price paths without mixing future outcomes into the input feature.
- Research question
- After unusually high relative volume, did prices tend to continue, reverse or show no reliable difference in a defined sample?
- What it cannot prove
- An eligible non-event observation is not a matched control; volume alone does not prove buying pressure or profitable execution.
| Published feature range (UTC) | 2021-08-01 → 2026-09-28. Endpoints do not prove continuous coverage. |
|---|---|
| Included symbols | AAVEUSDT · ADAUSDT · ARBUSDT · ATOMUSDT · AVAXUSDT · BCHUSDT · BNBUSDT · BTCUSDT · DOGEUSDT · DOTUSDT · ETCUSDT · ETHUSDT · FILUSDT · INJUSDT · LINKUSDT · LTCUSDT · NEARUSDT · OPUSDT · SOLUSDT · SUIUSDT · TRXUSDT · UNIUSDT · XLMUSDT · XRPUSDT |
| Rows | 726,239 features · 3,416,105 separate forward labels. Repeated horizons are not independent observations. |
| Bulk files | 107.3 MiB Parquet ZIP, plus an Excel/WPS-ready CSV ZIP split at 500,000 data rows per file. Both include coverage and opening instructions. |
Delivered feature fields: record_key, symbol, venue, event_time, observed_at, ingested_at, feature_time, source_available_at, historical_retrospective, method_version, valid_current, previous_gap, relative_quote_volume_24h, return_1h_pct, return_24h_pct
Inspect intervals, per-partition dates and rows in the bundled coverage report. Hourly/daily price features and native settlements must not be mixed without alignment.
Aggressive buying pressure
- What data is this?
- Spot taker-buy share, validity flags and price context for the supported markets, with separate forward outcomes where available.
- Why is it useful?
- Test whether aggressive buyers dominated executed spot trading before a move, while excluding rows where the buy-side measure is unavailable.
- Research question
- Did strong active buying coincide with a price move, and how did comparable valid observations behave afterwards?
- What it cannot prove
- Taker-buy share is executed-trade pressure, not exchange deposits, wallet inflows or a measure of all market demand.
| Published feature range (UTC) | 2021-08-01 → 2026-09-28. Endpoints do not prove continuous coverage. |
|---|---|
| Included symbols | AAVEUSDT · ADAUSDT · ARBUSDT · ATOMUSDT · AVAXUSDT · BCHUSDT · BNBUSDT · BTCUSDT · DOGEUSDT · DOTUSDT · ETCUSDT · ETHUSDT · FILUSDT · INJUSDT · LINKUSDT · LTCUSDT · NEARUSDT · OPUSDT · SOLUSDT · SUIUSDT · TRXUSDT · UNIUSDT · XLMUSDT · XRPUSDT |
| Rows | 726,239 features · 3,416,105 separate forward labels. Repeated horizons are not independent observations. |
| Bulk files | 107.3 MiB Parquet ZIP, plus an Excel/WPS-ready CSV ZIP split at 500,000 data rows per file. Both include coverage and opening instructions. |
Delivered feature fields: record_key, symbol, venue, event_time, observed_at, ingested_at, feature_time, source_available_at, historical_retrospective, method_version, valid_current, previous_gap, taker_buy_valid, taker_buy_share, return_1h_pct, return_24h_pct
Inspect intervals, per-partition dates and rows in the bundled coverage report. Hourly/daily price features and native settlements must not be mixed without alignment.
Price momentum
- What data is this?
- Backward-looking spot returns and trend features over multiple horizons, with later returns and path outcomes kept in separate label files.
- Why is it useful?
- Study trend continuation or reversal without accidentally using a future outcome as a historical decision input.
- Research question
- Did strong recent momentum continue over the next observation window after costs and risk assumptions are applied separately?
- What it cannot prove
- Reference-price outcomes are not executed profits; retrospective features do not prove they were available to a live strategy at the event time.
| Published feature range (UTC) | 2021-08-01 → 2026-09-28. Endpoints do not prove continuous coverage. |
|---|---|
| Included symbols | AAVEUSDT · ADAUSDT · ARBUSDT · ATOMUSDT · AVAXUSDT · BCHUSDT · BNBUSDT · BTCUSDT · DOGEUSDT · DOTUSDT · ETCUSDT · ETHUSDT · FILUSDT · INJUSDT · LINKUSDT · LTCUSDT · NEARUSDT · OPUSDT · SOLUSDT · SUIUSDT · TRXUSDT · UNIUSDT · XLMUSDT · XRPUSDT |
| Rows | 726,239 features · 3,416,105 separate forward labels. Repeated horizons are not independent observations. |
| Bulk files | 119.7 MiB Parquet ZIP, plus an Excel/WPS-ready CSV ZIP split at 500,000 data rows per file. Both include coverage and opening instructions. |
Delivered feature fields: record_key, symbol, venue, event_time, observed_at, ingested_at, feature_time, source_available_at, historical_retrospective, method_version, valid_current, previous_gap, bar_return_pct, trend_24h, return_1h_pct, return_4h_pct, return_24h_pct, return_168h_pct
Inspect intervals, per-partition dates and rows in the bundled coverage report. Hourly/daily price features and native settlements must not be mixed without alignment.
Volatility & path risk
- What data is this?
- Observed spot volatility, candle range and body measurements, paired with separate later return and favorable/adverse path labels where available.
- Why is it useful?
- Compare how much a trade idea could move for and against you after low- or high-volatility conditions, rather than looking only at its final return.
- Research question
- Did high-volatility setups have larger adverse excursions or different follow-through in a specified sample?
- What it cannot prove
- Favorable/adverse price paths are reference observations, not guaranteed fills, stop execution or net P&L.
| Published feature range (UTC) | 2021-08-01 → 2026-09-28. Endpoints do not prove continuous coverage. |
|---|---|
| Included symbols | AAVEUSDT · ADAUSDT · ARBUSDT · ATOMUSDT · AVAXUSDT · BCHUSDT · BNBUSDT · BTCUSDT · DOGEUSDT · DOTUSDT · ETCUSDT · ETHUSDT · FILUSDT · INJUSDT · LINKUSDT · LTCUSDT · NEARUSDT · OPUSDT · SOLUSDT · SUIUSDT · TRXUSDT · UNIUSDT · XLMUSDT · XRPUSDT |
| Rows | 726,239 features · 3,416,105 separate forward labels. Repeated horizons are not independent observations. |
| Bulk files | 113.7 MiB Parquet ZIP, plus an Excel/WPS-ready CSV ZIP split at 500,000 data rows per file. Both include coverage and opening instructions. |
Delivered feature fields: record_key, symbol, venue, event_time, observed_at, ingested_at, feature_time, source_available_at, historical_retrospective, method_version, valid_current, previous_gap, volatility_24h_pct, range_pct, body_pct, return_24h_pct
Inspect intervals, per-partition dates and rows in the bundled coverage report. Hourly/daily price features and native settlements must not be mixed without alignment.
Funding × OI × price events
- What data is this?
- Aligned historical funding percentiles, open-interest changes, spot-price context and joint event/completeness flags, with separate later outcomes where available.
- Why is it useful?
- Test specific funding-plus-OI crowding hypotheses using shared timestamps and explicit missing-input states instead of hand-joining separate histories.
- Research question
- What followed an extreme funding reading when open interest was also expanding, compared with eligible non-event observations?
- What it cannot prove
- Non-events are not automatically matched controls. Incomplete inputs and overlapping outcome windows require explicit exclusions or sensitivity checks.
| Published feature range (UTC) | 2021-08-01 → 2026-09-28. Endpoints do not prove continuous coverage. |
|---|---|
| Included symbols | AAVEUSDT · ADAUSDT · ARBUSDT · ATOMUSDT · AVAXUSDT · BCHUSDT · BNBUSDT · BTCUSDT · DOGEUSDT · DOTUSDT · ETCUSDT · ETHUSDT · FILUSDT · INJUSDT · LINKUSDT · LTCUSDT · NEARUSDT · OPUSDT · SOLUSDT · SUIUSDT · TRXUSDT · UNIUSDT · XLMUSDT · XRPUSDT |
| Rows | 683,221 features · 3,416,105 separate forward labels. Repeated horizons are not independent observations. |
| Bulk files | 100.9 MiB Parquet ZIP, plus an Excel/WPS-ready CSV ZIP split at 500,000 data rows per file. Both include coverage and opening instructions. |
Delivered feature fields: record_key, symbol, feature_time, historical_retrospective, method_version, price_reference_venue, return_24h_pct, volatility_24h_pct, funding_settled_at, funding_age_seconds, funding_percentile_90d, oi_source_time, oi_quantity_change_24h_pct, extreme_high_funding, extreme_low_funding, oi_expansion_5pct, oi_contraction_5pct, any_event, joint_inputs_complete, availability_assumption_seconds
Inspect intervals, per-partition dates and rows in the bundled coverage report. Hourly/daily price features and native settlements must not be mixed without alignment.
Relative strength & rotation
- What data is this?
- Historical 24-hour performance relative to BTC, cross-sectional percentiles and observed-market counts for the supported quote group.
- Why is it useful?
- Rank and compare which observed altcoins actually outperformed BTC on a given date while keeping the size of the observed comparison set visible.
- Research question
- Which coins were stronger than BTC in the observed set, and was that ranking based on broad enough coverage?
- What it cannot prove
- This is not a survivorship-free market universe or a recommendation to buy the top-ranked coin; missing or newly listed assets change the comparison set.
| Published feature range (UTC) | 2021-08-01 → 2026-09-28. Endpoints do not prove continuous coverage. |
|---|---|
| Included symbols | AAVEUSDT · ADAUSDT · ARBUSDT · ATOMUSDT · AVAXUSDT · BCHUSDT · BNBUSDT · BTCUSDT · DOGEUSDT · DOTUSDT · ETCUSDT · ETHUSDT · FILUSDT · INJUSDT · LINKUSDT · LTCUSDT · NEARUSDT · OPUSDT · SOLUSDT · SUIUSDT · TRXUSDT · UNIUSDT · XLMUSDT · XRPUSDT |
| Rows | 683,221 features · 0 separate forward labels. Repeated horizons are not independent observations. |
| Bulk files | 15.7 MiB Parquet ZIP, plus an Excel/WPS-ready CSV ZIP split at 500,000 data rows per file. Both include coverage and opening instructions. |
Delivered feature fields: record_key, symbol, venue, feature_time, method_version, quote_group, return_24h_pct, relative_to_btc_24h_pp, cross_section_percentile_24h, observed_markets, valid_markets, observed_positive_fraction, complete_historical_universe
Inspect intervals, per-partition dates and rows in the bundled coverage report. Hourly/daily price features and native settlements must not be mixed without alignment.
Historical recovery is retrospective, not evidence of what was known live at the time. Files stay usable for internal research after the 90-day hosted-access window. Future editions, raw records and redistribution rights are not included.
Which processed datasets are in the sealed archive?
The 2026-09-28 internal batch contains 30,057 completed candidate packages across the registered research families, plus audited settled-funding features for 24 contracts. These are packaged features, labels and context with methods, batch-level quality checks and file hashes. Package counts describe internal partitions, not customer entitlements or separate products.
Tell us the research question first: funding and OI regimes, price paths and risk, cross-market conditions, or public event outcomes. The table identifies the dataset that addresses it. Source row counts refer to the frozen inputs; processed feature rows have different grains and are confirmed for the selected delivery.
| Processed asset | Research use | Frozen source coverage | Delivery status |
|---|---|---|---|
| Settled funding features | Compare extreme settlements, rolling funding levels and consecutive positive or negative periods. | 120,684 source observations · 24 markets2021-12-01 → 2026-08-3124 internal candidate packages | Selected derived scope published above; raw/source remainder excluded |
| 5-minute open interest features | Separate position quantity from USDT notional changes and test their relationship with spot prices. | 10,605,126 source observations · 24 markets2021-12-01 → 2026-09-261,216 internal candidate packages | Selected derived scope published above; raw/source remainder excluded |
| Price, volume and forward paths | Segment trend, volatility and active-buy context; keep 1–168 hour future paths in separate labels. | 44,344,972 source observations · 391 markets2021-08-01 → 2026-09-2826,836 internal candidate packages | Selected derived scope published above; raw/source remainder excluded |
| Funding × OI × price observations | Study funding and position changes on one hourly grid alongside non-event observations. | Built from frozen upstream features938 internal candidate packages | Selected derived scope published above; raw/source remainder excluded |
| Market breadth and relative strength | Compare selected markets with BTC, ranks and the observed market universe. | Built from frozen upstream features62 internal candidate packages | Selected derived scope published above; raw/source remainder excluded |
| Spot–perpetual basis context | Inspect basis size and reject observations whose source times do not align. | 130,316 source observations2026-09-22 → 2026-09-28132 internal candidate packages | Separate scope and source-use review |
| Cross-venue market context | Compare venues and market types while checking the age of each observation. | 406,734 source observations2026-09-22 → 2026-09-28133 internal candidate packages | Separate scope and source-use review |
| Historical derivatives context | Study recorded OI and prices while preserving unknown funding-rate types. | 1,051,353 source observations2026-09-09 → 2026-09-2820 internal candidate packages | Separate scope and source-use review |
| Options term structure and skew | Compare expiry, 25-delta skew and butterfly measures where observed. | 20,586 source observations2026-09-22 → 2026-09-28132 internal candidate packages | Separate scope and source-use review |
| Coinbase active trade flow | Inspect observed buyer-initiated activity, buyer share and average trade size. | 61,698 source observations2026-09-06 → 2026-09-2823 internal candidate packages | Separate scope and source-use review |
| Hyperliquid OI samples | Check within-bucket OI changes and actual sampling coverage. | 852,264 source observations2026-09-11 → 2026-09-2818 internal candidate packages | Separate scope and source-use review |
| Observed liquidation events | Group observed events by contract and direction; coverage is not all exchange liquidations. | 209,792 source observations2026-09-06 → 2026-09-2823 internal candidate packages | Separate scope and source-use review |
| Stablecoin peg deviations | Compare observed dollar-peg deviations in basis points and flag large departures. | 37,945 source observations2026-09-09 → 2026-09-2820 internal candidate packages | Separate scope and source-use review |
| Listing announcements and revisions | Separate announcement, discovery and scheduled trading times; inspect later revisions. | 81,203 source observations2026-09-05 → 2026-09-2723 internal candidate packages | Separate scope and source-use review |
| Recorded public event index | Find public rule events and join them to separately stored later outcomes. | 7,441 source observations2026-09-05 → 2026-09-2824 internal candidate packages | Separate scope and source-use review |
| Public event outcome labels | Inspect later reference returns, path risk and missing or pending outcomes. | 37,185 source observations2026-09-05 → 2026-09-2824 internal candidate packages | Separate scope and source-use review |
| Research snapshot index | Find retained research versions and numeric paths; supporting metadata only. | 25,004 source observations2026-09-11 → 2026-09-2818 internal candidate packages | Separate scope and source-use review |
| Source snapshot index | Trace retained source values, market membership and observation time. | 142,996 source observations2026-09-11 → 2026-09-28397 internal candidate packages | Separate scope and source-use review |
| Collector health history | Explain collection delays and gaps; internal quality support. | 128,296 source observations2026-09-11 → 2026-09-2818 internal candidate packages | Internal quality support |
What does one processed row represent?
Price and OI features use one observation per market and timestamp. Other context families can use a record-key × metric long table, with a role identifying a feature or supporting metadata. Counts across these grains are not directly comparable.
Future paths belong in a separate labels table. Join only on the documented record key and keep future labels out of model inputs.
What accompanies a reviewed package?
The internal package contract includes features.parquet, labels.parquet, schema.json, method.md, provenance.json, quality.json, audit.json and manifest.json. Some families have no eligible outcomes and an empty labels table.
A sample.csv contains at most the first 100 feature rows; it demonstrates fields and nulls, not representative strategy performance. Private paths must be replaced by public source identifiers before any customer delivery.
Batch checks establish file integrity and calculation audit. Historical backfills were computed later; an event date does not show the value was available at that time. The published pack catalogue above lists the selected derived products available as one-time file downloads with optional API retrieval. The original sealed candidate batch remains private: raw source archives and unlisted families are not included. Inspect the public samples and per-release schemas before purchase.
Historical funding and open interest: does your contract fit?
Compare both fields for the same contract. Funding dates below describe imported settled history; they do not describe current funding quotes. Funding and OI endpoints can differ.
24 of 24 contracts
| Binance USD-M contract | Settled funding · rows / UTC dates | 5m open interest · rows / UTC dates |
|---|---|---|
| AAVEUSDT | 5,2052021-12-01 → 2026-08-31 | 427,5472022-09-03 → 2026-09-26 |
| ADAUSDT | 5,2052021-12-01 → 2026-08-31 | 427,5472022-09-03 → 2026-09-26 |
| ARBUSDT | 3,7732023-03-23 → 2026-08-31 | 369,4762023-03-23 → 2026-09-26 |
| ATOMUSDT | 5,2052021-12-01 → 2026-08-31 | 427,5472022-09-03 → 2026-09-26 |
| AVAXUSDT | 5,2052021-12-01 → 2026-08-31 | 427,2002022-09-04 → 2026-09-26 |
| BCHUSDT | 5,2052021-12-01 → 2026-08-31 | 427,2212022-09-04 → 2026-09-26 |
| BNBUSDT | 5,2052021-12-01 → 2026-08-31 | 507,0272021-12-01 → 2026-09-26 |
| BTCUSDT | 5,2052021-12-01 → 2026-08-31 | 507,0282021-12-01 → 2026-09-26 |
| DOGEUSDT | 5,2052021-12-01 → 2026-08-31 | 507,0282021-12-01 → 2026-09-26 |
| DOTUSDT | 5,2052021-12-01 → 2026-08-31 | 427,2592022-09-04 → 2026-09-26 |
| ETCUSDT | 5,2052021-12-01 → 2026-08-31 | 427,1222022-09-04 → 2026-09-26 |
| ETHUSDT | 5,2052021-12-01 → 2026-08-31 | 507,0282021-12-01 → 2026-09-26 |
| FILUSDT | 5,2052021-12-01 → 2026-08-31 | 427,2592022-09-04 → 2026-09-26 |
| INJUSDT | 4,4292022-08-16 → 2026-08-31 | 427,2582022-09-04 → 2026-09-26 |
| LINKUSDT | 5,2052021-12-01 → 2026-08-31 | 427,2582022-09-04 → 2026-09-26 |
| LTCUSDT | 5,2052021-12-01 → 2026-08-31 | 427,2152022-09-04 → 2026-09-26 |
| NEARUSDT | 5,2052021-12-01 → 2026-08-31 | 426,9702022-09-04 → 2026-09-26 |
| OPUSDT | 4,6582022-06-01 → 2026-08-31 | 427,2592022-09-04 → 2026-09-26 |
| SOLUSDT | 5,2802021-12-01 → 2026-08-31 | 507,0102021-12-01 → 2026-09-26 |
| SUIUSDT | 3,6492023-05-03 → 2026-08-31 | 357,3692023-05-03 → 2026-09-26 |
| TRXUSDT | 5,2052021-12-01 → 2026-08-31 | 427,2512022-09-04 → 2026-09-26 |
| UNIUSDT | 5,2052021-12-01 → 2026-08-31 | 427,2502022-09-04 → 2026-09-26 |
| XLMUSDT | 5,2052021-12-01 → 2026-08-31 | 426,9702022-09-04 → 2026-09-26 |
| XRPUSDT | 5,2052021-12-01 → 2026-08-31 | 507,0272021-12-01 → 2026-09-26 |
OI source checks (updated 2026-09-28): 37,472 imported daily files; 151 contain missing expected 5-minute slots. This separately updated file-level check is not a completeness percentage for the frozen batch. Request a contract-specific gap report.
Historical candle coverage by interval
These are stored candle records, not a claim that every interval covers every market from the earliest date. Derived resampling and exact symbol-level continuity require a separate check.
| Interval | Markets | Rows | Earliest UTC date | Latest UTC date |
|---|---|---|---|---|
| 1m | 36 | 1,972,442 | 2025-09-11 | 2026-09-28 |
| 5m | 391 | 35,129,680 | 2025-09-01 | 2026-09-28 |
| 15m | 34 | 31,859 | 2026-09-04 | 2026-09-28 |
| 1h | 341 | 6,815,978 | 2021-08-01 | 2026-09-28 |
| 4h | 25 | 3,180 | 2026-08-17 | 2026-09-27 |
| 1d | 339 | 391,833 | 2021-09-01 | 2026-09-27 |
What does one legacy event-study observation include?
Identity and known context
Row ID, symbol, venue, event type, observation time and calculation version. Price, volume, usable active-buy data and available derivative context retain their validity and provenance flags.
Historical imports are retrospective. A historical event timestamp alone does not prove the features were known then.
Subsequent price outcomes
15m, 1h, 4h, 24h and 7d reference-price windows, with completeness status and available favorable/adverse excursions. Pending and incomplete outcomes remain missing, not zero.
Hourly comparison rows are not automatically matched controls. Future outcomes are labels, never contemporaneous model inputs.
Existing authorized legacy customers retain their applicable API and JSON/CSV/Parquet access. This is separate from the ten one-time packs above, not an additional entitlement with every new purchase. Preserve release ID, schema and checksums. Reference returns are not executed profits and exclude execution costs.
Legacy evaluation sample: 2026-09-25-v2. Selected examples demonstrate fields and quality states, not representative strategy performance.
Questions before you choose
What does a purchased pack contain?
The complete published edition in two formats: CSV ZIP for Excel/WPS and Parquet ZIP for code, with opening instructions, field dictionary, per-partition coverage and checksums. Browser downloads need no API key.
Are all internal archives included?
No. Only the named, reviewed pack scope is included. Raw archives and unlisted families are excluded; interval endpoints do not imply uninterrupted coverage.
Do the packs contain live point-in-time data?
These historical features are retrospectively calculated. Their recovery timestamps must not be presented as original live availability.
Confirm your research window before buying.
Choose a one-time research data pack or the ten-pack bundle on the pricing page. Each purchase covers the named fixed edition and its published fields, markets and dates—not every internal partition. Hosted downloads and optional API retrieval last 90 days; downloaded files remain usable for permitted internal research afterwards. No recurring subscription or automatic updates are included.
Need a different scope? Send contracts, venues, dates, intervals, fields and intended use for a separately quoted coverage and permitted-use review. Free current Research access and existing legacy subscriptions remain separate.