← Back to Documentation

Data Sources & Validation

One cross-venue view of liquidation exposure, with clear source coverage, consistent field definitions, and checks you can repeat in your own workflow.

Methodology updated: October 9, 2026.

Supported sources and coverage

The position map brings together Hyperliquid, GMX, Gains Network, Drift, and Injective. Each venue has its own collection method, described below. Your subscription and request determine the included venues, symbols, and distance range.

Venue	Position discovery	Liquidation-price method
Hyperliquid	Leaderboard-wallet discovery with incremental, rate-limited rescans.	Reads `liquidationPx` from the wallet's clearinghouse state.
GMX	Open positions through a subgraph.	Calculated from position and collateral inputs using the collector's maintenance-margin assumptions.
Gains Network	Open trades through the backend feed.	Calculated from open price, leverage, and a liquidation threshold.
Drift	Sampled user accounts via RPC; not all accounts.	Approximation using position inputs and USDC collateral. Other collateral and cross-position account effects can change the venue's actual liquidation boundary.
Injective	Open positions from the venue feed.	Reads the feed's liquidation-price field with unit conversion.

Coverage is scoped to the supported sources and collected accounts, rather than the entire crypto market. Range filters, scan timing, pagination, and subscription settings determine the population shown.

Compare tier access

Reported inputs versus calculated fields

Entry price: the normalized API returns null when the entry price is unavailable, rather than presenting the current price as a real entry.
USD position value: generally calculated as native size multiplied by entry price, with venue-specific fallback paths and rounding. This defines the valuation basis for comparison across sources.
Unrealized P&L: calculated from current price, entry price, size, and side. Entry-dependent P&L fields are null if entry is unavailable. This arithmetic is not the venue's full accounting of funding or fees.
Collateral: uses reported collateral where available; otherwise the normalized API derives it from USD value and leverage.
Distance: calculated as the absolute difference between liquidation price and the effective mark price, divided by that mark price, expressed as a percentage.
Heatmap bands: group position exposure by price level. Exposure near a possible liquidation price is different from a liquidation event that has already occurred.

Display normalization applies venue-specific leverage limits. Use the field definitions above when comparing records with a source venue.

Read the API reference

How to validate a snapshot

These checks help you evaluate the data for your workflow and make comparisons reproducible.

Record context. Save the UTC retrieval time, symbol, venue filters, tier, distance limit, and revision identifiers. Keep raw responses so results can be reproduced.
Walk every page. Compare unique record IDs and counts across the complete result set. A first page is not a full snapshot, and live data may change during pagination.
Reconcile totals. Compare position sums and heatmap totals only for matching populations and filters, allowing for rounding and collection-time differences.
Inspect field relationships. Compare size, entry price, USD value, P&L, and leverage using the field definitions. Check both individual records and aggregate totals.
Check against venue state. Compare sampled records with the originating venue. For calculated liquidation prices, evaluate assumptions rather than expecting equality with a venue-reported field.
Compare candles. Match symbol, candle interval, timestamp, completed-candle status, and volume units before comparing OHLCV with an independent provider.

Combine record-level venue comparisons with distribution checks and total reconciliation for a well-rounded evaluation.

Track changes during an evaluation

The authenticated /api/v1/version endpoint exposes revision identifiers and notes:

measurement_revision identifies measurement-related changes. Read the associated notes and their affects_data flags to determine which dataset or output is affected.
tier_config_revision tracks API, feature, and configuration changes. Monitor it for changes that may affect your integration.
data_revision is a deprecated alias for measurement_revision. New integrations should use the explicit field.

Store the field name and relevant notes alongside its value. The counters have different meanings: do not compare them as if they shared one number space. Revision counters are not per-record freshness indicators.

Timing and interpretation

Venues are collected on their own schedules. Use matching time windows when comparing responses.
Timestamp definitions: observed_utc records the collector scan time when supplied, or the normalization time otherwise. timestamp_utc records normalization time.
Source confidence: a source-quality indicator, not a statistical accuracy percentage.
Calculated liquidation prices: interpret these using the venue-specific methods above, especially for cross-margin and multi-collateral accounts.
Dated comparisons: retain the test date and scope so subsequent measurements can be compared consistently.
ML and cascade outputs are experimental analytical inputs, not guaranteed forecasts or profitable trading signals.

Verify independently before using data in a trading decision. When reporting a discrepancy, include the endpoint, UTC time, filters, revisions, and affected record IDs. Never include passwords or access tokens.