Research-ready crypto data. Choose how you access it.

Research articles and current datasets

PUBLIC NOW. Published Research articles are free to read. After signing in, a free account can create one API key to query all current Research datasets as JSON or CSV at up to 60 requests per minute. Public page downloads remain available, but neither route includes historical Market Events releases.

Market Events & Outcomes

SELF-SERVICE NOW. The current fixed-plan dataset. Query versioned event/control rows through the customer API or download daily JSON, CSV and Parquet partitions.

Institutional & custom datasets

SCOPED DELIVERY. Derivatives, market regime, flow, listings, universe history and historical bars can be scoped only after coverage, source-use, fields, dates and delivery requirements are checked.

Live market research

Public current data. Data: Market breadth and regime, Bitcoin dominance and altcoin rotation, volatility and volume rankings, funding/OI/basis/spread views, liquidation pressure, spot-flow confirmation, correlations, trending assets and event calendars. Actual coverage: Each Research page declares its own venues, universe, refresh interval, observation time and methodology. There is no single implied all-market scope. Suitable for: Current market screening, source-backed editorial research, AI citation and deciding whether a deeper historical dataset is worth scoping. Delivery: Public Research pages plus a free customer API key for current JSON/CSV datasets, limited to 60 requests per minute per account. Important boundary: Current snapshots are not a historical API, raw exchange payload or fixed-plan bulk download.

Market Events & Outcomes

Self-service product. Data: Recorded price, volume, RSI, MACD, EMA, Bollinger and aggressive-flow events; hourly comparison rows; 5-minute context; available derivatives context; 15m/1h/4h/24h/7d returns, MFE/MAE and quality flags. Actual coverage: Market Events v1 currently covers six specified Binance USDT spot pairs and only the published event-study history shown in its catalog. Suitable for: Signal-family evaluation, participation filters, event/comparison studies and reproducible research-agent workflows. Delivery: Researcher and Research Desk customer API; immutable daily JSON, CSV and Parquet with schema, manifest and checksums. Important boundary: This is the only dataset automatically included in the current $99 and $399 fixed plans.

Derivatives positioning & stress

Scope before purchase. Data: Venue-labelled funding, open interest and changes, perpetual basis, liquidation events and pressure, crowding and squeeze components with direction evidence and missing-state labels. Actual coverage: Internal observations are accumulating across supported derivatives sources. Historical depth, contract mapping, native funding semantics and usable fields differ by venue and period. Suitable for: Crowding studies, leverage-regime analysis, liquidation-event research and spot-versus-perpetual confirmation. Delivery: Custom extract or institutional API/files only after an actual coverage and commercial-source review. Important boundary: Not included automatically in fixed plans; not a complete exchange-wide tick/liquidation archive or settled funding ledger.

Market regime, liquidity & flow

Scope before purchase. Data: Breadth, sector strength, relative volume, CVD and validated aggressive trading, realised volatility/ATR, trend strength, correlations, order-book liquidity snapshots and cross-venue spot confirmation. Actual coverage: Universe, venues and historical depth vary by measure. Trade-side flow is executed buying/selling, not deposits, withdrawals or blockchain fund flows. Suitable for: Regime classification, rotation analysis, liquidity screens, volatility filters, cross-market confirmation and research feature construction. Delivery: Current public snapshots now; historical derived files or a task-specific dataset only after field and coverage review. Important boundary: A current dashboard value does not imply a continuous historical series or point-in-time-complete universe.

Cross-market & on-chain context

Public current / scoped history. Data: Coinbase-versus-Binance spot premium proxies, venue volume share and price dislocation, DEX volume by chain, stablecoin supply context and cross-venue confirmation. Actual coverage: Current derived snapshots from named venues and aggregators. Quote assets, chain definitions, refresh intervals and available history differ by measure; no single metric represents all crypto capital flows. Suitable for: Check whether a move is confirmed across venues, compare centralized and decentralized activity, and add liquidity or stablecoin context to market research. Delivery: Public current Research pages now; historical or task-specific derived files only after field, date, source-use and coverage review. Important boundary: Not wallet-level on-chain flows, deposits or withdrawals, institutional order flow, a complete chain history or a guaranteed arbitrage price. Premium is a cross-market proxy.

Listings & scheduled market events

Scope before purchase. Data: Verified listings and delistings, planned/changed/open trading states, token unlock observations and post-listing checkpoints with event and observation timestamps. Actual coverage: Official exchange or named-source observations from CoinNudge's actual capture period. Announcement, scheduled opening and confirmed trading are kept as different states. Suitable for: Listing-event studies, announcement-to-open timelines, post-listing observation and scheduled-event research. Delivery: Public current event pages; custom event files or institutional delivery after source, sample size and date coverage checks. Important boundary: Not included in Market Events v1. Sparse samples do not support a universal listing success rate.

Point-in-time universe & asset identity

Accumulating / scoped. Data: Observed daily market membership, venue/symbol identity mapping, asset classification and exclusions, liquidity thresholds, sector membership and source-health context. Actual coverage: Only states actually observed after local capture began can be point-in-time evidence. Pre-capture historical membership is not reconstructed from today's coin list. Suitable for: Reducing survivorship bias, reproducing an eligible universe and explaining why tokenized equities, stable-value assets or illiquid markets were excluded. Delivery: Institutional or custom delivery after confirming the requested date, venue and classification fields. Important boundary: Still accumulating and not included in fixed plans; it is not proof of every venue's complete historical product directory.

Historical market bars

Backfill in progress / scoped. Data: Closed Binance USDT spot OHLCV at selected 1m, 5m, 1h and 1d intervals, including valid aggressive-buy amounts where available, with event/observation/ingestion time semantics and coverage checks. Actual coverage: Targets and actual completion differ by symbol and timeframe; long-history backfill is still running. Current-universe backfill can carry survivorship bias, and backfilled rows were not known locally at their historical event time. Suitable for: Return, volatility, volume and intraday-structure research where retrospective market bars are acceptable and gaps are explicitly handled. Delivery: Custom bulk Parquet/CSV only after checking the exact symbol/timeframe/date coverage and source-use eligibility. Important boundary: Not part of current fixed plans, not a raw trade/order-book product, and not automatically valid for point-in-time backtesting.

What data are you actually buying?

Market Events v1 is a structured event-study dataset, not a raw exchange feed. One row represents a recorded public signal or a scheduled hourly comparison observation. The same row connects its identity, measured context, later price behaviour and quality flags.

Event identity & trigger

Categorical labels + numeric trigger metrics. Seven recorded public signal families: price, volume, RSI, MACD, EMA, Bollinger and aggressive flow. Values and thresholds retain the original rule units; they are not comparable probabilities. Delivered fields: row_id, symbol, venue, event_kind, direction, event_timeframe, trigger_value, trigger_threshold. How you can use it: Group events by coin, direction and rule family. These are recorded triggers, not continuous RSI/MACD indicator histories or proof that every possible trigger was captured.

Observation & availability times

UTC Unix timestamps + age in seconds. Separate source event time, local recording time, the chosen candle close and when its features became available. sample_time is the event recording time or scheduled control time. Delivered fields: sample_time, event_time, event_recorded_at, price_time, feature_available_at, feature_age_seconds. How you can use it: Align a study to an explicit observation clock and identify late inputs instead of assuming all values were available when an event occurred.

Price & trading activity

Closed-candle features: USDT amounts + ratio. Selected closed 5-minute price, executed quote volume and volume divided by the mean of the previous 20 contiguous 5-minute bars. Prefer the latest locally known eligible candle within 30 minutes; expose its age. Missing baseline means a missing ratio. Delivered fields: price, quote_volume_5m, volume_ratio_20. How you can use it: Compare events under stronger versus ordinary participation. This is a feature snapshot, not a full historical OHLCV candle file or executable quote.

Aggressive buying & selling

Validated USDT amounts + fraction from 0 to 1. Aggressive-buy share is validated buy quote volume divided by total quote volume. Net aggressive volume is twice buy quote volume minus total quote volume. Invalid amounts remain null. Delivered fields: taker_buy_quote_5m, taker_buy_fraction, net_taker_quote_5m, taker_status. How you can use it: Study whether buyer-led participation coincides with different post-event outcomes. These are executed trade-side measures, not deposits, withdrawals or wallet capital flows.

Optional perpetual context

Nullable Hyperliquid observation + provenance. Recent locally known Hyperliquid context only where eligible, within 30 minutes of the observation. OI is USD notional; funding, where populated, uses normalized hourly percentage points. The native funding type/settlement is not established and is explicitly flagged; do not use it as a settled funding payment or arbitrage quote. Missing observations remain null. Delivered fields: funding_hourly_pct, oi_usd, derivative_venue, derivative_time, derivative_available_at, derivative_status. How you can use it: Explore spot events alongside available leverage context after checking coverage. Not Binance contract data, not a guaranteed complete funding/OI history and not net inflow.

Scheduled comparison rows

Control observations using the same feature columns. Hourly observations for the six coins, excluding a recorded qualifying public event in the preceding inclusive hour. Controls carry the same context and forward-window structure. Delivered fields: sample_type=control, sample_time, control_status. How you can use it: Describe how event periods differ from scheduled comparison periods. These are not matched controls, randomized experiments or confirmed negative rule evaluations; overlapping samples need care.

Five forward outcome windows

Numeric percentage outcomes + completeness flags. Repeated for 15m, 1h, 4h, 24h and 7d. Raw return measures reference-price change; direction-adjusted return signs it by event direction. MFE/MAE are nonnegative favourable/adverse excursions under the recorded long/short convention. Controls have no direction-adjusted return. Delivered fields: return_pct_*, direction_adjusted_return_pct_*, mfe_pct_*, mae_pct_*, status_*, missing_candles_*, expected_candles_*. How you can use it: Compare subsequent direction, magnitude and adverse movement at several horizons. The reference is the first full subsequent 5-minute interval, not an actual trade. Pending/gap windows stay missing; fees and slippage are not modelled.

Research eligibility & reproducibility

Quality labels, booleans and release metadata. Distinguish locally known, retrospective and missing context. An immutable release includes its field dictionary, coverage and checksums; updated data receives another release ID. Delivered fields: calculation_version, feature_status, feature_pit_eligible; release ID, schema, manifest, SHA-256. How you can use it: Apply explicit inclusion rules and repeat an analysis on the same version. Feature eligibility alone is not proof of a complete point-in-time universe or an unbiased backtest.

Signal-family evaluation

For Independent quants and strategy researchers. Question: How do recorded MACD, EMA or volume events differ over 1h, 4h and 24h? Method: Group event rows by family, coin and direction; require complete selected windows. Report counts, median reference returns and adverse-excursion distributions. Separate overlapping observations and validate outside the development sample. Possible output: A signal-family comparison table with counts and quality exclusions—not a claim of a profitable strategy. Value: Starts with recorded triggers already connected to context and outcome columns, reducing repeated event-to-candle alignment work.

Participation-filter research

For Quant analysts and research notebooks. Question: Do volume events accompanied by stronger aggressive buying have different outcomes? Method: Use volume events, volume_ratio_20 and valid taker_buy_fraction. Compare declared groups within the same coin/direction; exclude missing flow and require locally-known features for at-the-time predictor tests. Possible output: Group sizes, missingness and forward-outcome comparisons. Do not select a threshold on the same sample and call it validated. Value: Combines validated trade-side features and future labels so research can focus on the hypothesis rather than joining feeds.

Event versus comparison periods

For Academic and institutional exploratory research. Question: Is the observed post-event movement unusual relative to scheduled comparison periods? Method: Compare event and control rows by coin and observation period using compatible raw-return windows. Account for direction conventions, overlapping windows, timing and sample imbalance; controls are not automatically matched. Possible output: A descriptive event/control comparison with an explicit study design and limitations, not a causal estimate. Value: Supplies comparison observations as well as selected events, making it possible to examine more than the triggered sample alone.

Reproducible reports and agent workflows

For Research desks, analysts and external research agents. Question: Can another researcher trace this chart or conclusion back to the exact rows? Method: Pin a daily release ID, retain schema and checksums, filter rows through the API or load Parquet locally, and report dates, formula version, exclusions and counts with every analysis. Possible output: A traceable report or notebook with a recorded input version. The analyst or agent still performs and checks the analysis. Value: Provides versioned evidence for reruns and review instead of relying on a changing dashboard screenshot.

What the subscription does—and does not—replace

You pay for organized event records, aligned features, comparison observations, forward labels and versioned delivery—not exclusive access to public exchange prices or a proven investment edge. You still choose the hypothesis, assess sample adequacy, control bias, model transaction costs and validate the result. Short capture history may support exploration but not a statistically reliable strategy conclusion.

One daily file is not several years of history

Download means the complete selected UTC day's published event/control rows. It does not mean all internally backfilled multi-year candles. The API queries the same versioned rows; it is not a live tick stream. Published partitions are checked approximately hourly, so API and files have the same data freshness. Only actually published dates are included; inspect coverage before purchase. Raw trades, full order books, continuous historical indicator series, multi-year OHLCV exports, listing datasets and the entire internal Research archive are not included in this fixed-plan product. Private customer records are never part of delivery. A custom dataset needs separate scope and availability confirmation.

Choose delivery by research workload: JSON REST API or CSV/Parquet files

Both Researcher and Research Desk include the same Market Events v1 dataset and JSON/CSV/Parquet formats, with different key allowances, account limits and personal/team use. Query an immutable daily release by symbol and event/control type, or download its files, manifest, schema and checksums. Custom extracts require separate scope confirmation, not an assumed fixed-plan entitlement. Prices are shown only on the pricing page.

Signal events and forward observations

One row represents a qualifying public event or a scheduled hourly control on six fixed spot pairs. Controls exclude recorded qualifying events in the prior hour; this is not proof of no rule trigger or complete capture. Context uses closed candles, valid taker amounts and locally known Hyperliquid observations where available. Retrospective feature flags prevent late/backfilled observations from masquerading as then-known data. Forward 15m/1h/4h/24h/7d returns and excursions are future labels, not predictors, executions or profit claims.

What questions can this dataset support?

Investigate whether recorded volume signals behave differently with stronger aggressive-buy participation, compare events with hourly comparison observations on the same coin, or describe favorable and adverse price moves by signal family. Start with sample coverage: short history, late inputs, overlapping observations and missing derivatives context can make a study unsuitable. These are descriptive comparisons, not causal estimates or a profitable strategy.

Five parts of one dataset—not five separate products

Each row identifies a public directional price, volume, RSI, MACD, EMA, Bollinger or aggressive-flow event, or a scheduled control. Columns describe the observation, comparison status, closed-candle context, five forward outcomes and quality. MFE means maximum favorable reference-price movement; MAE means maximum adverse movement. Missing or pending outcomes are null, not zero. The first full subsequent 5-minute candle supplies the reference start; no execution or fee model is included.

A row needs provenance, not just a number

Source event time, event recording time and feature availability time are separate fields. Retrospective inputs arrived after the observation time and must not be treated as then-known information. Selected price context has a timestamp and age. A complete 24-hour outcome does not make its original inputs point-in-time eligible. This is not a historical all-market universe or a complete point-in-time database.

Commercial data boundary

The fixed plans include Market Events v1, not the full internal archive, all Research datasets, listings datasets or a historical all-market universe. Exchange-native payload resale, raw tick feeds, full order books, private accounts, payments, Telegram identities and personal alerts are outside scope. Trade execution and guaranteed performance are not provided; external redistribution requires separate review.

Sample → plan → key → dataset

Inspect actual rows and the full dictionary at /data/samples. Choose a data plan on /pricing?product=data, sign in and complete payment. Then open API Keys, confirm the data entitlement, create an independent customer API key and copy it once. Pick a published daily release to download or query. Save its release ID and schema for reproducibility. Data subscriptions are separate from personal alert access.

Updates and available history

The recent nine UTC days are checked hourly; changed daily partitions receive new release IDs. Older releases remain available, but older gaps are not guaranteed to be repaired automatically. Each manifest reports actual dates, event/control counts and outcome completeness. The public preview is a small real sample, not the full paid history; no fixed historical row count is promised.

Published research

• Subscribe to data
• View real sample
• Customer API Keys