HIP-4 Market Activity Report: August 2026

What does the August 2026 HIP-4 snapshot show?

Purrdict’s August 3 snapshot contains 1,818 HIP-4 outcome-token rows, $403,183,505.4047749 in cumulative indexed notional, and 5,723,936 indexed fills. Activity was not dominated by a handful of rows: the top 10 outcome-token rows accounted for 14.45% of indexed notional, while the top 100 accounted for 48.74%.

The other standout is activity density. The 72 recurring price-binary rows were only 3.96% of the archive’s rows, yet they produced 23.99% of indexed notional and 35.96% of fills.

These figures describe records in Purrdict’s independent HIP-4 index. They are not claims about current liquidity, open interest, unique questions, or protocol-wide unique traders. The dated JSON and CSV snapshot is public under CC BY 4.0 so the calculations can be reproduced.

Snapshot at a glance

The snapshot was generated on August 3, 2026 UTC from Purrdict’s indexed market-overview endpoint.

Outcome-token rows do not represent unique questions. A binary question generally produces two token rows, and a multi-outcome question produces more. The archive therefore avoids presenting 1,818 as a prediction-market count.

Finding 1: indexed notional is dispersed across token rows

The largest outcome-token row contributed 2.57% of cumulative indexed notional. Concentration rose gradually rather than immediately:

This means no single outcome-token row explains the archive. It does not prove that economic activity is equally distributed across unique questions. Related token rows can belong to the same question, and incomplete historical metadata prevents a reliable question-level grouping for the whole snapshot.

The distinction matters for analysts. Ranking individual #N tokens is reproducible from the public file; collapsing them into questions requires a separate, validated relationship layer.

Finding 2: recurring price binaries punch above their row count

The parser identified 72 recurring price-binary rows. They were 3.96% of rows but accounted for:

• $96,736,630.55150996, or 23.99% of indexed notional;
• 2,058,553 fills, or 35.96% of indexed fills;
• approximately 28,591 fills per outcome-token row.

That is the highest fill density among the parsed classes in this snapshot. It suggests recurring bounded-price questions have generated repeated trading activity within the indexed history.

It does not establish why. A single cumulative snapshot cannot separate product design, listing duration, market visibility, volatility, incentive effects, or user preference. Answering those questions requires time-series data and question-level metadata.

Finding 3: named outcomes and price binaries carry much of classified activity

Named outcomes and recurring price binaries together comprise 212 rows—11.66% of the archive—but contribute 43.08% of indexed notional and 57.58% of fills.

The result should be read as a parser-coverage finding, not a complete protocol taxonomy. The unclassified group still contains more than half of indexed notional.

Finding 4: metadata coverage is the largest analytical constraint

Purrdict could defensibly classify 427 rows, or 23.49% of the archive. The other 1,391 rows remain unclassified because the indexed description was missing, incomplete, or did not match a known shape.

Keeping those rows is important. Dropping them would remove $216,664,945.74324852 in indexed notional and 2,162,357 fills, making the classified part of the archive look more complete than it is. Guessing a class would create a different problem: a neat chart built on unverifiable labels.

For builders, the practical lesson is to preserve raw HIP-4 identifiers and source metadata before adding a presentation taxonomy. A UI can call a token “Yes,” “Argentina,” or a price range, but the underlying #N identity must remain available for reconciliation and order construction. The HIP-4 protocol guide explains why identifier context matters.

How the analysis was calculated

The Purrdict HIP-4 market-data repository contains the generator, tests, methodology, dated snapshot, and latest distributions. This report uses snapshots/2026-08-03/overview.json and performs four transparent operations:

1. Sum volume across all rows for cumulative indexed notional.
2. Sum trades across all rows for indexed fill count.
3. Group rows by market_class, retaining blank values as unclassified.
4. Sort rows by volume and divide cumulative top-row volume by total volume for concentration shares.

No missing period is estimated. No row is removed because its description is unclassified. Per-row trader fields cannot be summed into a protocol-wide unique-trader count because the same address can trade multiple outcome tokens.

The complete data landing page exposes the live JSON endpoint, latest JSON snapshot, latest CSV snapshot, and the broader limitations.

What this snapshot cannot answer

• Current liquidity: cumulative notional does not reveal today’s spread, order-book depth, or executable size.
• Growth: one dated snapshot cannot establish whether activity is rising or falling.
• Unique questions: outcome-token rows are not a question count.
• Unique traders: addresses overlap across rows.
• Complete class shares: 76.51% of rows are not yet classified from available descriptions.
• Trading performance: historical average prices and volume do not measure returns or predict future results.
• Protocol completeness: totals cover successfully indexed records, not a guaranteed lossless account of every HIP-4 event.

For a live order, use Hyperliquid’s current market data and inspect the book at the intended size. The archive is a research and integration resource, not an execution quote.

What Purrdict will measure next

Future dated snapshots can turn this baseline into time-series research. The next useful measures are changes in indexed notional and fills by class, concentration over time, newly observed identifiers, classification coverage, and the difference between cumulative activity and current executable depth.

Each future report should retain the same unit definitions. Changing “outcome-token row” into “market” between periods would create a growth number that cannot be reproduced.

Builders can inspect the open files today, use @purrdict/hip4 for tested identifier and order utilities, or use mute.sh when they need normalized historical tables and streaming delivery. Traders should use Purrdict’s live interface for current HIP-4 books.

Purrdict is an independent application and data publisher built on public Hyperliquid infrastructure. It is not operated by or affiliated with Hyperliquid Labs.